Methods, apparatus, and system for wireless detection, sensing, and learning

Abstract

To provide methods, an apparatus and systems for wireless detection, sensing and learning.SOLUTION: A described method comprises: transmitting a wireless signal through a wireless multipath channel of a venue; receiving the wireless signal through the wireless multipath channel, where the received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel that is impacted by a motion of an object in the venue; obtaining a time series of channel information (TSCI) of the wireless multipath channel; computing an autocorrelation function (ACF) based on the TSCI; computing at least one ACF feature of the ACF or a function of the ACF; computing an ACF statistics, a motion statistics, a speed statistics, and a gait statistics based on the at least one ACF feature; and in response to a determination that the motion is detected, classifying the object associated with the detected motion as a human or a non-human, based on: the ACF statistics, the motion statistics, the speed statistics, and the gait statistics.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present teachings relate generally to wireless detection, sensing, and learning. More specifically, the present teachings relate to wireless sensing and proximity detection of objects relative to targets within a venue, wireless detection and discrimination of human and non-human activity, and deep learning-based wireless sensing. [Background technology]

[0002] The demand for human proximity detection is increasing significantly in a variety of indoor applications. Human proximity detection refers to the ability to identify and monitor the presence and proximity of individuals relative to a specific area, object, or system. This technology plays a key role in home automation, indoor localization, security, and efficiency improvement in a variety of situations. Common vision-based technologies such as cameras, infrared, and radar are often used for human motion detection and location, but they can invade privacy or require specialized sensors for accurate measurements. When additional sensors are unavailable or privacy is a concern, alternative solutions are needed.

[0003] As the popularity of Internet of Things (IoT) devices skyrockets, these devices increasingly use WiFi connectivity, and wireless sensing is finding a major role to play in IoT applications. By leveraging the interaction of WiFi radios during signal propagation, wireless devices can passively detect human motion and activity through radio signal analysis. This approach makes passive indoor proximity detection more practical and achievable. Instead of relying on invasive vision-based techniques, human presence and motion can be detected non-intrusively by analyzing radio signal propagation. This advancement in WiFi-based proximity sensing enables more privacy-conscious solutions that can be used in practical applications.

[0004] However, implementing a robust proximity detector using on-device WiFi signals is not a trivial task. WiFi signals' ability to penetrate walls and other obstacles, while useful for whole-home coverage, hinders proximity detection. While some research has explored the use of WiFi signals for motion detection, it has primarily focused on motion detection within a wide coverage area and overlooked the specific aspect of proximity detection. Other research has focused on localization, but accurate localization often requires complex setups involving multiple transceivers with specific geometric arrangements, as well as dedicated calibration and training. Such complexity is unnecessary for proximity detection.

[0005] With the proliferation of Internet of Things (IoT) devices, indoor intelligent applications such as security monitoring, intruder detection, occupancy monitoring, and behavior recognition have attracted significant attention. However, these applications often suffer from high false alarm rates due to their inability to distinguish between human and non-human objects, such as pets, robotic vacuum cleaners, and electrical appliances. This discrimination ability is essential, especially for applications related to security, health monitoring, automation, and energy management. Misrecognition leads to user frustration and erodes trust, hindering the practical application and widespread adoption of these technologies. Given the prevalence of pets, robotic vacuum cleaners, and electrical appliances, especially in residential environments, it is crucial to develop reliable systems that can accurately recognize human and non-human objects.

[0006] Accurately distinguishing between various human and non-human activities remains a challenge today. For example, camera-based and thermal sensor-based approaches can only detect moving objects within line-of-sight (LOS). Furthermore, camera-based systems raise privacy concerns. Strategies that utilize radar to distinguish between pets and humans based on vital signs expect pets to remain stationary, which is unrealistic. Additionally, these methods have strict device placement requirements, are often limited to line-of-sight, and assume that objects are moving within a given area. Summary of the Invention

[0007] The present teachings relate generally to wireless detection, sensing, and learning. More specifically, the present teachings relate to wireless sensing and proximity detection of objects relative to targets within a venue, wireless detection and discrimination of human and non-human activity, and deep learning-based wireless sensing.

[0008] In one embodiment, a system for wireless sensing is described. The system includes a first wireless device, a second wireless device, and a processor. The first wireless device is configured to transmit a wireless signal through a wireless multipath channel in a venue. The second wireless device is configured to receive the wireless signal through the wireless multipath channel. The received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel and movement of an object in the venue. The processor is configured to obtain a time series of channel information (TSCI) of the wireless multipath channel based on the received wireless signal, calculate first operation statistics based on a time-domain correlation of the channel information (CI) in the TSCI, calculate second operation statistics based on a frequency-domain correlation of the CI in the TSCI, and detect an approach of an object to a target in the venue based on the first operation statistics and the second operation statistics.

[0009] In another embodiment, a method of wireless sensing is described, the method including: transmitting, by a first wireless device, a wireless signal over a wireless multipath channel in a venue; receiving, by a second wireless device, the wireless signal over the wireless multipath channel, where the received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel and a motion of an object in the venue; obtaining a time series of channel information (TSCI) of the wireless multipath channel based on the received wireless signal; calculating first operation statistics based on a time-domain correlation of the channel information (CI) in the TSCI; calculating second operation statistics based on a frequency-domain correlation of the CI in the TSCI; and detecting proximity of the object to a target in the venue based on the first operation statistics and the second operation statistics.

[0010] In yet another embodiment, an apparatus for wireless sensing is described. The apparatus includes a receiver configured to receive a wireless signal transmitted by a transmitter over a wireless multipath channel in a venue, a memory having an instruction set stored therein, and a processor communicatively coupled to the memory and the receiver. The received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel and movement of an object within the venue. The processor is configured to obtain a time series of channel information (TSCI) of the wireless multipath channel based on the received wireless signal, calculate first operating statistics based on a time-domain correlation of the channel information (CI) in the TSCI, calculate second operating statistics based on a frequency-domain correlation of the CI in the TSCI, and detect proximity of an object to a target within the venue based on the first operating statistics and the second operating statistics.

[0011] In another embodiment, a system for wireless human-non-human motion detection is described. The system includes a transmitter configured to transmit a wireless signal over a wireless multipath channel of a venue, a receiver configured to receive the wireless signal over the wireless multipath channel of the venue, and a processor. The received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel being affected by motion of an object within the venue. The processor obtains a time series of channel information (TSCI) of the wireless multipath channel based on the received wireless signal, calculates an autocorrelation function (ACF) based on the TSCI, calculates at least one ACF feature of the ACF or a function of the ACF, calculates ACF statistics based on the ACF or a first ACF feature of the ACF function, calculates motion statistics related to object motion based on the ACF or a second ACF feature of the ACF function, calculates speed statistics related to speed of the object motion based on the ACF or a third ACF feature of the ACF function, and calculates gait statistics related to walking motion of the object based on the ACF or a fourth ACF feature of the ACF function, and, in response to determining that motion has been detected, classifies the object associated with the detected motion as either human or non-human based on the ACF statistics, the motion statistics, the speed statistics, and the gait statistics.

[0012] In another embodiment, a method for wireless human-non-human motion detection is described, which includes transmitting a wireless signal through a wireless multipath channel of a venue, receiving the wireless signal through the wireless multipath channel of the venue, where the received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel being affected by the motion of an object within the venue, obtaining a time series of channel information (TSCI) of the wireless multipath channel based on the received wireless signal, calculating an autocorrelation function (ACF) based on the TSCI, calculating at least one ACF feature of the ACF or a function of the ACF, and calculating a first AC feature of the ACF or a function of the ACF. calculating ACF statistics based on the F features; calculating motion statistics associated with the object's motion based on a second ACF feature of the ACF or a function of the ACF; calculating speed statistics associated with the speed of the object's motion based on a third ACF feature of the ACF or a function of the ACF; and calculating gait statistics associated with the object's walking motion based on a fourth ACF feature of the ACF or a function of the ACF; and in response to determining that motion has been detected, classifying the object associated with the detected motion as either human or non-human based on the ACF statistics, motion statistics, speed statistics, and gait statistics.

[0013] In yet another embodiment, an apparatus for wireless human-non-human motion detection is described. The apparatus includes a receiver configured to receive a wireless signal transmitted by a transmitter over a wireless multipath channel within a venue, a memory having an instruction set stored therein, and a processor communicatively coupled to the memory and the receiver. The received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel being affected by the motion of an object within the venue. The processor is configured to obtain a time series of channel information (TSCI) of the wireless multipath channel based on the received wireless signal, calculate an autocorrelation function (ACF) based on the TSCI, calculate at least one ACF feature of the ACF or a function of the ACF, calculate ACF statistics based on the ACF or a first ACF feature of the ACF function, calculate motion statistics related to a motion of the object based on the ACF or a second ACF feature of the ACF function, calculate speed statistics related to a speed of the motion of the object based on a third ACF feature of the ACF or a function of the ACF, and calculate gait statistics related to a walking motion of the object based on a fourth ACF feature of the ACF or a function of the ACF, and in response to determining that motion has been detected, classify the object associated with the detected motion as either human or non-human based on the ACF statistics, the motion statistics, the speed statistics, and the gait statistics.

[0014] In another embodiment, a system for wireless sensing is described. The system includes a transmitter configured to transmit a wireless signal over a wireless multipath channel of a venue, a receiver configured to receive the wireless signal over the wireless multipath channel of the venue, and a processor. The wireless multipath channel is affected by the motion of an object within the venue. The received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel and the motion of the object. The processor is configured to obtain a time series of channel information (TSCI) of the wireless multipath channel based on the received wireless signal, calculate multiple autocorrelation functions based on the TSCI, each autocorrelation function (ACF) being calculated based on a CI of the TSCI in a respective sliding time window, construct at least one ACF vector, each ACF vector being a vector associated with a respective ACF including multiple vector elements associated with a respective time lag, each vector element being a value of a respective ACF evaluated at a respective time lag, rearrange the at least one ACF vector into rearranged ACF data, each ACF vector being a one-dimensional (1D) ACF block, and perform a wireless sensing task based on a task engine that processes using the rearranged ACF data as input.

[0015] In another embodiment, a method for wireless sensing is described. The method includes transmitting a wireless signal through a wireless multipath channel of a venue, the wireless multipath channel being affected by a motion of an object within the venue; receiving the wireless signal through the wireless multipath channel of the venue, the received wireless signal differing from the transmitted wireless signal due to the wireless multipath channel and the motion of the object; obtaining a time series of channel information (TSCI) of the wireless multipath channel based on the received wireless signal; calculating a plurality of autocorrelation functions (ACFs) based on the TSCI, each autocorrelation function (ACF) being calculated based on a CI of the TSCI in a respective sliding time window; constructing at least one ACF vector, each ACF vector being a vector associated with a respective ACF including a plurality of vector elements associated with a respective time lag, each vector element being a value of a respective ACF evaluated at a respective time lag; rearranging the at least one ACF vector into rearranged ACF data, each ACF vector being a one-dimensional (1D) ACF block; and performing a wireless sensing task based on a task engine that processes using the rearranged ACF data as input.

[0016] In yet another embodiment, an apparatus for wireless sensing is described. The apparatus includes a receiver configured to receive a wireless signal transmitted through a wireless multipath channel of a venue, and a processor. The wireless multipath channel is affected by the motion of an object within the venue. The received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel and the motion of the object. The processor is configured to obtain a time series of channel information (TSCI) of the wireless multipath channel based on the received wireless signal, calculate multiple autocorrelation functions based on the TSCI, each autocorrelation function (ACF) calculated based on a CI of the TSCI in a respective sliding time window, construct at least one ACF vector, each ACF vector being a vector associated with a respective ACF including multiple vector elements associated with a respective time lag, each vector element being a value of a respective ACF evaluated at a respective time lag, rearrange the at least one ACF vector into rearranged ACF data, each ACF vector being a one-dimensional (1D) ACF block, and perform a wireless sensing task based on a task engine that processes the rearranged ACF data as input.

[0017] Other concepts relate to software for implementing the present teachings of wireless detection, sensing, and learning. Additional novel features will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following description and the accompanying drawings, or may be learned by the manufacture or operation of the embodiments. The novel features of the present teachings may be realized and attained by practice or use of various aspects of the methodology, apparatus, and combinations that are described in the detailed embodiments that follow. [Brief explanation of the drawings]

[0018] The methods, systems, and / or devices described herein will be further explained in terms of exemplary embodiments, which will be described in detail with reference to the drawings, which are non-limiting exemplary embodiments, and in which like reference numerals represent like structure throughout the several views of the drawings.

[0019] [Figure 1] FIG. 1 is a diagram of an exemplary system for proximity detection, according to some embodiments of the present disclosure.

[0020] [Figure 2] 10 illustrates several example correlation matrices between subcarriers at different human-to-device distances, in accordance with some embodiments of the present disclosure.

[0021] [Figure 3] 1 illustrates gait information from difference of autocorrelation functions (ACFs), according to some embodiments of the present disclosure.

[0022] [Figure 4] FIG. 1 illustrates a finite state machine (FSM) for proximity monitoring, according to some embodiments of the present disclosure.

[0023] [Figure 5] FIG. 1 illustrates a process including different states regarding the proximity of an object to a target, according to some embodiments of the present disclosure.

[0024] [Figure 6A] 1 illustrates an exemplary environment and deployment for wireless proximity detection, according to some embodiments of the present disclosure. [Figure 6B] 1 illustrates an exemplary environment and deployment for wireless proximity detection, according to some embodiments of the present disclosure. [Figure 6C] 1 illustrates an exemplary environment and deployment for wireless proximity detection, according to some embodiments of the present disclosure. [Figure 6D] 1 illustrates an exemplary environment and deployment for wireless proximity detection, according to some embodiments of the present disclosure.

[0025] [Figure 7A] 1 illustrates exemplary metrics for wireless proximity detection, according to some embodiments of the present disclosure. [Figure 7B] 1 illustrates exemplary metrics for wireless proximity detection, according to some embodiments of the present disclosure. [Figure 7C] 1 illustrates exemplary metrics for wireless proximity detection, according to some embodiments of the present disclosure.

[0026] [Figure 8] 1 illustrates a flowchart of an exemplary method for wireless proximity detection, according to some embodiments of the present disclosure.

[0027] [Figure 9] 10 is a flowchart illustrating detailed operations of wireless proximity detection according to some embodiments of the present disclosure.

[0028] [Figure 10] FIG. 1 illustrates interference of non-human actions with a human detection system, according to some embodiments of the present disclosure.

[0029] [Figure 11] 1 illustrates movement patterns of a human, a dog, and a cleaning robot, according to some embodiments of the present disclosure.

[0030] [Figure 12] FIG. 1 is a diagram of an exemplary system for wireless human-non-human motion detection, according to some embodiments of the present disclosure.

[0031] [Figure 13A] 1 illustrates CSI, ACF, motion statistics, and speed estimation for a human, a pet, and a cleaning robot, according to some embodiments of the present disclosure. [Figure 13B] 1 illustrates CSI, ACF, motion statistics, and speed estimation for a human, a pet, and a cleaning robot, according to some embodiments of the present disclosure. [Figure 13C] 1 illustrates CSI, ACF, motion statistics, and speed estimation for a human, a pet, and a cleaning robot, according to some embodiments of the present disclosure.

[0032] [Figure 14A] 10A-10C illustrate histograms of motion statistics and speed estimation for humans and pets, respectively, according to some embodiments of the present disclosure. [Figure 14B] 10A-10C illustrate histograms of motion statistics and speed estimation for humans and pets, respectively, according to some embodiments of the present disclosure.

[0033] [Figure 15] 1 illustrates an example of a human stride cycle, according to some embodiments of the present disclosure.

[0034] [Figure 16A] 10A-10C illustrate examples of stride cycle times and stride lengths for humans and pets, according to some embodiments of the present disclosure. [Figure 16B] 10 illustrates examples of stride cycle times and stride lengths for humans and pets, according to some embodiments of the disclosure.

[0035] [Figure 17A] 25 and 75 percentiles of velocity estimation for a human, a robot, and a pet, respectively, according to some embodiments of the present disclosure. [Figure 17B] 25 and 75 percentiles of velocity estimation for a human, a robot, and a pet, respectively, according to some embodiments of the present disclosure.

[0036] [Figure 18A] 1 illustrates the ACF, average ACF peak value, and average ACF valley value at a time instance for a human, a robot, and a pet, according to some embodiments of the present disclosure. [Figure 18B]1 illustrates the ACF, average ACF peak value, and average ACF valley value at a time instance for a human, a robot, and a pet, according to some embodiments of the present disclosure. [Figure 18C] 1 illustrates the ACF, average ACF peak value, and average ACF valley value at a time instance for a human, a robot, and a pet, according to some embodiments of the present disclosure.

[0037] [Figure 19] 1 illustrates an example of motion detection with motion statistics and amplified ACF segmentation, according to some embodiments of the present disclosure.

[0038] [Figure 20] 1 illustrates an exemplary process of data input preparation for human-non-human action recognition using deep learning, according to some embodiments of the present disclosure.

[0039] [Figure 21] 1 illustrates a flowchart of an exemplary method for wireless human-non-human motion detection, according to some embodiments of the present disclosure.

[0040] [Figure 22] 1 is a flowchart illustrating detailed operation of wireless human-non-human motion detection according to some embodiments of the present disclosure.

[0041] [Figure 23] 1 illustrates a flowchart of an exemplary method for wireless sensing, according to some embodiments of the present disclosure.

[0042] [Figure 24] 1 is an example block diagram of a first wireless device of a system for wireless detection, sensing, and / or learning, according to some embodiments of the present disclosure.

[0043] [Figure 25]FIG. 10 is an example block diagram of a second wireless device of a system for wireless detection, sensing, and / or learning, in accordance with some embodiments of the present disclosure. DETAILED DESCRIPTION OF THE INVENTION

[0044] The symbol " / " disclosed herein means "and / or." For example, "A / B" means "A and / or B." In some embodiments, a method / device / system / software for a wireless monitoring system is disclosed. Time-series channel information (CI) of a wireless multipath channel is obtained using a processor, a memory communicatively coupled to the processor, and a set of instructions stored in the memory. The time-series CI (TSCI) may be extracted from a wireless (sounding) signal transmitted through the channel from a Type 1 heterogeneous wireless device (e.g., a wireless transmitter (TX), a "Bot" device) to a Type 2 heterogeneous wireless device (e.g., a wireless receiver (RX), an "Origin" device) within a venue. The channel is affected by the representation / motion of objects within the venue. The characteristics / spatial-temporal information (STI) / motion information (MI) of the objects / representations / motions may be calculated / monitored based on the TSCI. A task (e.g., sensing objective) can be performed based on the characteristics / STI / MI. A task-related presentation can be generated in a user interface (UI) on the user's device. The characteristics / STI / MI can be a motion index value (MIV) / activity data (AD) / motion data (MD) / motion sensing data (MSD) / motion detection data (MDD) / motion score (MS) / motion statistics (MS2).

[0045] Expressions may include placement, arrangement of moving parts, location / speed / acceleration / position / orientation / direction / identifiable place / area / presence / spatial coordinates, static expression / presentation / state / size / length / width / height / angle / scale / curve / surface / area / volume / pose / attitude / embodiment / body language, dynamic expression / motion / sequence / movement / action / gesture / stride / extension / contraction / distortion / deformation, bodily expression (e.g., head / face / eyes / mouth / tongue / hair / voice / neck / limbs / arms / hands / feet / muscles / moving parts), surface expression / shape / texture / material / color / electromagnetic (EM) properties / visual pattern / wetness / reflection / translucency / flexibility, material properties (e.g., biological tissue / hair / cloth / metal / wood / leather / plastic / man-made materials / solid / liquid / gas / temperature), changes in expression, and / or any combination.

[0046] A wireless multipath channel may include a communication channel, an analog frequency channel (e.g., with a carrier frequency near 700 / 800 / 900 MHz or 1.8 / 1.8 / 1.9 / 2.4 / 3 / 5 / 6 / 27 / 60 / 70+ GHz), a coded channel (e.g., in CDMA), and / or a channel of a wireless / cellular network / system (e.g., WLAN, WiFi, mesh, 4G / LTE / 5G / 6G / 7G / 8G / 9G / 10G, Bluetooth, Zigbee, UWB, RFID, microwave). It may include multiple channels, which may be contiguous (e.g., adjacent / overlapping bands) or non-contiguous (e.g., non-overlapping bands, 2.4 GHz / 5 GHz). While a channel is used to transmit wireless signals and perform sensing measurements, data (e.g., TSCI / features / components / characteristics / STI / MI / analytical values / task outputs, auxiliary / non-sensing data / network traffic) may also be communicated / transmitted over the channel.

[0047] The wireless signal may include a sequence of probe signals, which may be EM radiation, radio frequency (RF), optical, band-limited, or baseband signals, licensed or unlicensed ISM band signals, MIMO transmissions, sensing transmissions, wireless, mobile, cellular, optical, network, mesh, downlink, uplink, unicast, multicast, or broadcast signals, which may comply with standards / protocols (e.g., WLAN, WWAN, WPAN, WBAN, international, national, industrial, de facto, IEEE, 802.11, 15, 16, WiFi, 802.11n, ac, ax, be, bf, 3G, 4G, LTE, 5G, 6G, 7G, 8G, 9G, 10G, 3GPP, Bluetooth, BLE, Zigbee, NFC, RFID, UWB, WiMax). The probe signal may include any of the following: protocol / standard / beacon / pilot / sounding / excitation / illumination / handshake / synchronization / reference / source / motion probe / detection / sensing / management / control / data / null-data / beacon / pilot / request / response / association / reassociation / disassociation / authentication / action / report / poll / announcement / extension / query / acknowledgement frames / packets / signals, and / or null-data-frame (NDP), RTS, CTS, QoS, CF-Poll, CF-Ack, block-acknowledgement, reference, training, or synchronization. It may include line-of-sight (LOS) or non-LOS components (or paths / links). Data may be embedded. The probe signal may be replaced (or embedded) by a data signal. Each frame / packet / signal may include a preamble / header / payload. It may include a training sequence, short (STF) / long (LTF) training field, L-STF / L-LTF / L-SIG / HE-STF / HE-LTF / HE-SIG-A / HE-SIG-B, and a channel estimation field (CEF). It may be used to transfer power wirelessly from Type 1 devices to Type 2 devices. The sounding rate of the signal may be adjusted to control the amount of power transferred.The probe signal may be transmitted in bursts.

[0048] The TSCI may be extracted / obtained (e.g., by an IC / chip) from a wireless signal at a layer of a Type-2 device (e.g., a layer of the OSI reference model, PHY / MAC / Data Link / Logical Link Control / Network / Transport / Session / Presentation / Application layer, TCP / IP / Internet / Link layer). It may be extracted from a received wireless / derived signal. It may include wireless sensing measurements obtained with a communication protocol (e.g., wireless / cellular communication standard / network, 4G / LTE / 5G / 6G / 7G / 8G / 9G / 10G, WiFi, IEEE802.11 / 11bf / 15 / 16). Each CI may be extracted from a probe / sounding signal and associated with a timestamp. The TSCI may be associated with the CI / sampling / sounding frequency / cycle start time / stop time / duration / amount. The motion detection / sensing signal may be recognized / identified based on the probe signal. TSCI can be stored / retrieved / accessed / pre-processed / processed / post-processed / conditioned / analyzed / monitored. TSCI / features / components / characteristics / STI / MI / analysis / task results can be communicated to an edge / cloud server / Type 1 / Type 2 / hub / data aggregator / another device / system / network.

[0049] A Type 1 / Type 2 device may include components (hardware / software) such as electronics / chips / integrated circuits (ICs) / RF circuits / antennas / modems / TX / RX / transceivers / RF interfaces (e.g., 2.4 / 5 / 6 / 27 / 60 / 70+ GHz radios / fronthaul radios / backhaul radios), networks / interfaces / processors / memory / modules / circuits / boards / software / firmware / connectors / structures / enclosures / housings / structures. It may be an access point (AP) / base station / mesh / router / repeater / hub / radio station / client / terminal / "Origin Satellite" / "Tracker Bot" and / or an Internet of Things (IoT) / appliance / wearable / accessory / peripheral / funny / amenity / gadget / vehicle / module / wireless-enabled / unicast / multicast / broadcast / node / hub / target / sensor / portable / mobile / cellular / communication / motion detection / source / destination / standards-compliant device. It may include additional attributes such as auxiliary functions / network connectivity / generic / brand / model / appearance / form / shape / color / material / specifications, etc. The above (e.g., components / device types / additional attributes) may be heterogeneous, as they may differ for different Type 1 (or Type 2) devices.

[0050] Type 1 / Type 2 devices may or may not be authenticated / associated / co-located. They may be the same device. Type 1 / Type 2 / portable / nearby / other devices, sensing / measurement sessions / links between them, and / or objects / representations / motions / characteristics / STI / MI / tasks may be associated with an identity / identification / identifier (ID), such as a UUID, associated / unassociated STA ID (ASID / USID / AID / UID). Type 2 devices may passively observe / monitor / receive wireless signals from Type 1 devices without establishing a connection (e.g., association / authentication / handshake) with the Type 1 device or requesting services from the Type 1 device. Type 1 / Type 2 devices may move along with the tracked object / other object.

[0051] A Type 1 (TX) device may function temporarily / sporadically / continuously / repeatedly / interchangeably / alternatively / simultaneously as a Type 2 (RX) device, and vice versa. A Type 1 device may be a Type 2 device. A device may function temporarily / sporadically / continuously / repeatedly / simultaneously as a Type 1 / Type 2 device. There may be multiple wireless nodes, each a Type 1 / Type 2 device. TSCI may be obtained between two nodes when they exchange / communicate wireless signals. Object characteristics / STI / MI may be monitored based on TSCI individually or multiple TSCIs together.

[0052] The motion / representation of an object can be actively monitored using a Type 1 / Type 2 device that moves with the object (e.g., a wearable device / automated guided vehicle / AGV), or passively monitored with a Type 1 / Type 2 device that does not move with the object (e.g., both devices are fixed).

[0053] Tasks may be performed using a reference / trained / initial database / profile / baseline that is trained / collected / processed / computed / sent / stored during the training phase. The database may be retrained / updated / reset.

[0054] A presentation may include UI / GUI / text / message / form / webpage / visual / image / video / graphic / animation / graphical / symbol / emoji / sign / color / shade / sound / music / speech / audio / mechanical / gesture / vibration / haptic presentation. Time series of characteristics / STI / MI / task results / other quantities may be displayed / presented in a presentation. Any computation may be performed / shared by a processor (or logic unit / chip / IC), Type 1 / Type 2, user, nearby, another device, local, edge, cloud server, hub, data, signal analysis subsystem, sensing initiator / responder, SBP initiator / responder, AP, or non-AP. A presentation may include monthly / weekly / daily / simplified / detailed / cross-sectional / small / large / form factor / color-coded / comparison / summary / web display, animation, audio announcement, repetitive motion, or another presentation related to the cyclical / recurring characteristics of the representation.

[0055] Multiple Type 1 (or Type 2) devices may interact with a Type 2 (or Type 1) device. Multiple Type 1 (or Type 2) devices may be synchronous / asynchronous and / or may use the same / different channel / sensing parameters / settings (e.g., sounding frequency / bandwidth / antenna). A Type 2 device may receive another signal from a Type 1 / another Type 1 device. A Type 1 device may transmit another signal to a Type 2 / another Type 2 device. The wireless signals transmitted (or received) by them may be sporadic / temporary / continuous / repeated / synchronous / simultaneous. They may operate independently / collaboratively. Their data (e.g., TSCI / features / characteristics / STI / MI / intermediate task results) may be processed / monitored / analyzed independently or jointly / collaboratively.

[0056] Any device may act based on some state / internal state / system state. Devices may communicate directly or through another / nearby / portable device / server / hub device / cloud server. Devices / systems may be associated with one or more users with associated settings. Settings may be selected / pre-programmed / changed / adjusted / modified / altered over time. Methods may be performed in the order shown or in another order. Steps may be performed in parallel / iteratively / repeatedly. Users may include humans / adults / elderly / males / females / young people / children / babies / pets / animals / living things / machines / computer modules / software. Steps / actions / processing may differ for different devices based on location / orientation / direction / role / roles / user-related characteristics / settings / available resources / bandwidth / power / network connection / hardware / software / processors / coprocessors / memory / battery life / antennas / directional antennas / power settings / device parameters / characteristics / conditions / situations / states. Any / all devices may be controlled / coordinated by a processor (e.g., Type 1 / Type 2 / nearby / portable / associated with another device / server / designated source), including devices physically within / on / attached to a common device.

[0057] A Type 1 (or Type 2) device may be capable of wirelessly coupling with multiple Type 2 (or Type 1) devices. The Type 1 (or Type 2) device may be caused / controlled to switch / establish a wireless coupling (e.g., association / authentication) from a Type 2 (or Type 1) device to another Type 2 (or another Type 1) device. The switching may be controlled by a server / hub device / processor / Type 1 device / Type 2 device. The wireless channels before and after the switching may be different. A second wireless signal may be transmitted between the Type 1 (or Type 2) device and a second Type 2 (or second Type 1) device through the second channel. A second TSCI of the second channel may be extracted / obtained from the second signal. The first / second signal, the first / second channel, the first / second Type 1 device, and / or the first / second Type 2 device may be identical / similar / co-located.

[0058] A Type 1 device may transmit / broadcast a wireless signal to multiple Type 2 devices without establishing a connection (association / authentication) with each individual Type 2 device. It may transmit to a specific / common MAC address, which may be the MAC address of several devices (e.g., dummy receivers). Each Type 2 device may adapt to a specific MAC address to receive the wireless signal. The specific MAC address may be associated with a venue, which may be recorded in an association table of an association server (e.g., a hub device). The venue may be identified by the Type 1 device / Type 2 device based on the wireless signal received at the specific MAC address.

[0059] For example, a Type 2 device may be moved to a new venue. A Type 1 device may be newly set up at a venue such that the Type 1 and Type 2 devices are unaware of each other. During setup, the Type 1 device may be instructed / guided / made / controlled (e.g., by a dummy receiver, hardware pin configuration / connection, saved configuration, local configuration, remote configuration, downloaded configuration, hub device, and / or server) to transmit wireless signals (e.g., a sequence of probe signals) to a specific MAC address. When powered on, the Type 2 device may scan for probe signals according to a table of MAC addresses (e.g., stored in a designated source, server, hub device, cloud server) that may be used to broadcast at different locations (e.g., different MAC addresses used for different venues such as a house / office / enclosure / floor / multi-store building / store / airport / mall / stadium / hall / station / subway / lot / area / zone / region / district / city / country / continent, etc.). When the Type 2 device detects a wireless signal transmitted to a specific MAC address, it can use the table to identify the venue.

[0060] A channel may be selected from a set of candidate / selectable / acceptable channels. Candidate channels may be associated with different frequency bands / bandwidths / carrier frequencies / modulations / wireless standards / coding / encryption / payload characteristics / networks / IDs / SSIDs / characteristics / settings / parameters. A particular MAC address / selected channel may be changed / adjusted / modified over time (e.g., according to a time table / rule / policy / mode / condition / situation / change). The selection / change may be based on availability / collisions / traffic patterns / co-channel / inter-channel interference / available bandwidth / random selection / pre-selection list / plan. This may be done by a server (e.g., a hub device). They may be communicated (e.g., to / from a Type 1 / Type 2 / hub / another device / local / edge / cloud server).

[0061] A wireless connection (e.g., association / authentication) may be established (e.g., using a signal handshake) between the Type 1 device and a nearby / portable / another device. The Type 1 device may send a first handshake signal (e.g., a sounding frame / probe signal / request-to-send RTS) to the nearby / portable / another device. The nearby / portable / another device may respond to the first signal by sending a second handshake signal (e.g., a command / clear-to-send / CTS) to the Type 1 device, triggering the Type 1 device to transmit / broadcast a wireless signal to multiple Type 2 devices without establishing a connection with the Type 2 device. The second handshake signal may be a response / acknowledgement (e.g., ACK) to the first handshake signal. The second handshake signal may include information of the venue / Type 1 device. The nearby / portable / another device may be a dummy device having a purpose (e.g., primary purpose, secondary purpose) of establishing a wireless connection with the Type 1 device to receive the first signal or to transmit the second signal. The nearby / portable / another device may be physically attached to the Type 1 device.

[0062] In another example, a nearby / portable / other device may trigger the Type 1 device to broadcast a signal to multiple Type 2 devices without establishing connections with the multiple Type 2 devices by sending a third handshake signal to the Type 1 device, which may respond to the third signal by sending a fourth handshake signal to the other device.

[0063] A nearby / portable / other device may be used to trigger multiple Type 1 devices to broadcast. It may have multiple RF circuits to trigger multiple transmitters in parallel. The triggering may be sequential / partially sequential / partially / fully parallel. Parallel triggering may be achieved by using additional (one or more) devices to perform similar triggering in parallel with the nearby / portable / other device. After establishing a connection with the Type 1 device, the nearby / portable / other device may suspend / stop communication with the Type 1 device. It may enter an inactive / hibernate / sleep / standby / low power / off / power down mode. The suspended communication may be resumed. The nearby / portable / other device may have a specific MAC address, and the Type 1 device may send a signal to the specific MAC address.

[0064] A (first) wireless signal may be transmitted by a first antenna of a Type 1 device over a first channel at a first venue to several first Type 2 devices. A second wireless signal may be transmitted by a second antenna of a Type 1 device over a second channel at a second venue to several second Type 2 devices. The first and second signals may be transmitted at first and second (sounding) rates, respectively, possibly individually to first and second MAC addresses. Several first and second channels / signals / rates / MAC addresses / antennas / Type 2 devices may be the same / different / synchronous / asynchronous. The first and second venues may have the same / different size / shape / multipath characteristics. The immediate areas around the first and second venues / first and second antennas may overlap. The first / second channels / signals can be WiFi+LTE (one WiFi and one LTE), or WiFi+WiFi, or WiFi(2.4GHz)+WiFi(5GHz), or WiFi(5GHz, channel=a1, BW=a2)+WiFi(5GHz / channel=b1, BW=b2). Some of the first / second items (e.g., channel / signal / rate / MAC address / antenna / Type 1 / Type 2 device) can be changed / adjusted / modified over time (e.g., based on a time table / rule / policy / mode / condition / situation / another change).

[0065] Each Type 1 device may be a signal source for multiple Type 2 devices (i.e., it transmits separate probe signals to separate Type 2 devices). Each separate Type 2 device may asynchronously select a Type 1 device from among all Type 1 devices as its signal source. The TSCI may be obtained by each separate Type 2 device from a separate series of probe signals from the Type 1 device. A Type 2 device may (e.g., initially) select a Type 1 device from among all Type 1 devices as its signal source based on the Type 1 / Type 2 device's identity / identification / identifier, task, past signal source, history, characteristics, signal strength / quality, threshold for switching signal source, and / or user / account / profile / access information / parameters / inputs / requirements / criteria information.

[0066] A database of available / candidate Type 1 (or Type 2) devices may be initialized / maintained / updated by a Type 2 (or Type 1) device. A Type 2 device may receive wireless signals from multiple candidate Type 1 devices. It may select its Type 1 device (i.e., signal source) based on any of the following: signal quality / strength / regularity / channel / traffic / characteristics / attributes / conditions / task requirements / training task results / MAC address / identity / identifier / past signal sources / history / user instructions / other considerations.

[0067] An undesirable / bad / problematic / unsatisfactory / unacceptable / unacceptable / faulty / severe / undesirable / insufficient / insufficient / poor / unsuitable condition may occur when (1) the timing between adjacent probe signals in the received wireless signal becomes irregular and deviates from the agreed-upon sounding rate (e.g., time perturbations beyond an acceptable range) and / or (2) the processed / signal strength of the received signal is too weak (e.g., below a third threshold or below a fourth threshold for a significant percentage of the time), where processing includes any low-pass / band-pass / high-pass / median / moving / weighted average / linear / nonlinear / smoothing filtering. Any thresholds / percentages / parameters may vary over time. Such conditions may occur when Type 1 / Type 2 devices become increasingly distant or when the channel becomes congested.

[0068] Some settings (e.g., Type 1-Type 2 device pairing / signal source / network / association / probe signal / sounding rate / method / channel / bandwidth / system state / TSCI / TSMA / task / task parameters) may be changed / modified / adjusted / modified. The changes may be according to time tables / rules / policies / modes / conditions (e.g., undesirable conditions) / other changes. For example, the sounding rate may be normally 100Hz, but may be changed to 1000Hz in harsh conditions or to 1Hz in low power / standby conditions.

[0069] Settings can vary based on task requirements (e.g., 100Hz normal, 1000Hz instantaneously for 20 seconds). Within a task, instantaneous systems can be adaptively / dynamically associated with classes / states / conditions (e.g., low / normal / high priority / emergency / critical / normal / privileged / non-subscription / subscription / paid / non-paid). Settings (e.g., sounding rate) can be adjusted accordingly. Changes can be controlled by the server / hub / Type 1 / Type 2 device. Scheduled changes can be made according to a timetable. Changes can be made immediately if an emergency is detected or gradually if an ongoing condition is detected.

[0070] Characteristics / STIs / MIs may be monitored / analyzed individually based on TSCIs associated with a particular Type 1 / Type 2 device pair, jointly based on multiple TSCIs associated with multiple Type 1 / Type 2 pairs, jointly based on any TSCIs associated with a particular Type 2 device and any Type 1 device, jointly based on any TSCIs associated with a particular Type 1 device and any Type 2 device, or globally based on any TSCIs associated with any Type 1 / Type 2 device.

[0071] Classifiers / classification / recognition / detection / estimation / projection / feature extraction / processing / filtering may be applied (e.g., to CIs / CI features / characteristics / STIs / MIs) and / or trained / retrained / updated. In the training stage, training may be performed based on multiple training TSCIs of several training wireless multipath channels or characteristics / STIs / MIs calculated from the training TSCIs. The training TSCIs are obtained from training wireless signals transmitted from a training type 1 device and received by a training type 2 device. Retraining / updating may be performed in the operating stage based on the training TSCIs / current TSCIs. There can be multiple classes (e.g., groupings / categories / events / motions / representations / activities / objects / locations) attached to venue / region / zone / place / environment / house / office / building / warehouse / facility objects / representations / motions / actions / processes / events / manufacturing / assembly line / maintenance / repair / navigation / objects / emotions / minds / states / stages / gestures / gates / actions / motions / presence / motions / routines / activities / history / events.

[0072] Classifiers can include linear / nonlinear / binary / multi-class / Bayesian classifier / Fisher linear discriminant / logistic regression / Markov chain / Monte Carlo / deep / neural networks / perceptron / self-organizing maps / boosting / meta-algorithms / decision trees / random forests / genetic programming / kernel learning / KNN / support vector machines (SVM).

[0073] Feature extraction / projection may include any of the following: subspace projection / principal component analysis (PCA) / independent component analysis (ICA) / vector quantization / singular value decomposition (SVD) / eigendecomposition / eigenvalue / time / frequency / orthogonal / non-orthogonal decomposition, processing / pre-processing / post-processing. Each CI may contain multiple components (e.g., vectors / combinations of complex values). Each component may be pre-processed to give magnitude / phase or functions thereof.

[0074] Features can include: feature extraction / projection output, amplitude / amplitude / phase / energy / power / intensity / intensity, presence / absence / proximity / likelihood / histogram, time / period / duration / frequency / component / decomposition / projection / band, local / global / maximum / minimum / zero crossing, repeating / periodic / typical / habitual / transient / atypical / sudden / mutually exclusive / evolutionary / transitional / changing / temporal / associated / correlated features / patterns / trends / profiles / events / trends / slope / behavior, causality / short term / long term / correlation / statistics / frequency / period / duration, movement / movement / location / map / coordinates / height / speed / acceleration / angle / rotation / size / volume, suspicious / hazard / alarm event / warning / certainty / approach / collision Bump, tracking / breathing / heart rate / gait / behavior / event / statistics / hourly / daily / weekly / monthly / yearly parameters / statistics / analysis, health / disease / medical statistics / analysis, state / condition / situation / disease / biometric early / immediate / concurrent / delayed indication / suggestion / sign / indicator / verification / detection / symptom, baby / patient / machine / equipment / temperature / vehicle / parking lot / road / lift / elevator / space / street / fluid flow / house / room / office / home / building / warehouse / storage / system / ventilation / fan / pipe / duct / person / human / car / boat / truck / plane / drone / downtown / crowd / impulsive event / cyclostationary / environment / vibration / material / surface / 3D / 2D / local / global, and / or another measurable quantity / variable. Features may include monotonic functions of features or sliding aggregates of features within a sliding window.

[0075] Training can include AI / machine / deep / supervised / unsupervised / discriminative training / autoencoder / linear discriminant analysis / regression / clustering / tagging / labeling / Monte Carlo operations.

[0076] A current event / motion / expression / object at a current time may be classified by applying a classifier to current TSCI / characteristics / STI / MI obtained from current wireless signals received by a Type 2 device at a venue from a Type 1 device during the operational phase. If there are multiple Type 1 / Type 2 devices, some / all (or their locations / antenna locations) may be permutations of the corresponding training Type 1 / Type 2 devices (or locations / antenna locations). The Type 1 / Type 2 devices / signals / channels / venues / objects / motions may be the same as / different from the corresponding training entities. The classifier may be applied in a sliding window. The current TSCI / characteristics / STI / MI may be augmented by training TSCI / characteristics / STI / MI (or fragments / extractions) to bootstrap the classification / classifier.

[0077] A first section / segment (having a first duration / start / end time) of a first TSCI (associated with a first Type 1-Type 2 device pair) may be aligned with a second section / segment (having a second duration / start / end time) of a second TSCI (associated with a second Type 1-Type 2 device pair) (e.g., using dynamic time warping / DTW / matched filtering, possibly based on some mismatch / distance / similarity score / cost, or correlation / autocorrelation / cross-correlation), and each CI in the first section is mapped to a CI in the second section. The first / second TSCIs may be pre-processed. Some similarity scores (per component / item / link / segment) may be computed. The similarity scores may include any of the mismatch / distance / similarity score / cost. A component-wise similarity score may be calculated between a component of a first item (CI / feature / characteristic / STI / MI) in the first section and a corresponding component of a corresponding mapped item (second item) in the second section. An item-wise similarity score may be calculated between the first and second items (e.g., based on an aggregation of corresponding component-wise similarity scores). The aggregation may include a sum / weighted sum, a product / weighted product, a weighted average / robust / trimmed mean / arithmetic / geometric / harmonic mean, a median / mode, a percentile, or another aggregation of any of the above. A link-wise similarity score may be calculated between the first and second items associated with a link (TX-RX antenna pair) of the first and second Type 1-Type 2 device pair (e.g., based on an aggregation of corresponding item-wise similarity scores). A segment-wise similarity score may be calculated between the first and second segments (e.g., based on an aggregation of corresponding link-wise similarity scores). The first / second segment may be slid.

[0078] In DTW, the first / second segment, the first / second item, another first (or second) item of the first (or second) segment, or any function of the corresponding timestamp / duration / difference / difference may satisfy the constraint. The time difference between the first item and the second item may be constrained (e.g., upper / lower bound). The first (or second) section may be the entire first (or second) TSCI. The duration / start / end time of the first / second may be the same / different.

[0079] In one example, the first / second Type 1-Type 2 device pair may be the same, and the first / second TSCI may be the same or different. If different, the first / second TSCI may include a current / reference, current / current, or reference / reference TSCI pair. In the case of "current / reference," the first TSCI may be a current TSCI acquired during the operational phase, and the second TSCI may be a reference TSCI acquired during the training phase. In the case of "reference / reference," the first / second TSCI may be two TSCIs acquired during the training phase (e.g., for two training events / states / classes). In the case of "current / current," the first / second TSCI may be two TSCIs acquired during the operational phase (e.g., associated with two different antennas or two measurement setups). In another example, the first / second Type 1-Type 2 device pair may be different but share a common device (Type 1 or Type 2).

[0080] The aligned first / second segments (or portions thereof) can be represented as first / second vectors. The portions may include all items (for "segment-wise"), or all items associated with the TX-RX link (for "link-wise"), or items (for "item-wise"), or components of items (for "component-wise"). The similarity score may include any combination / aggregation / function of the following: dot product / correlation / autocorrelation / correlation index / covariance / discrimination score / distance / Euclidean / absolute / L_k / weighted distance (between the first / second vectors). The similarity score may be normalized by the vector length. Parameters derived from the similarity score may be modeled using statistical distributions. The scale / location / other parameters of the statistical distribution may be estimated.

[0081] Recall that there may be multiple sliding segments. A classifier may be applied to the sliding first / second segment pair to obtain a provisional classification result. It may associate the current event with a particular class based on one segment pair / provisional classification result or multiple segment pairs / provisional classification results (e.g., the period for which the similarity score is dominant (e.g., largest / smallest / dominant / unmatched / most important / superior), or is sufficiently significant (e.g., higher / lower than some threshold) among all candidate classes for N consecutive times, or for a high / low sufficient percentage, or the most / lowest frequent).

[0082] Channel information (CI) may include any of the following: signal strength / amplitude / phase / timestamp / spectral power measurements / modem parameters / dynamic beamforming information / beamforming reports / dynamic imaging information / channel display information / spatial maps from dynamic beamforming / transfer function components / radio conditions / measurable variables / sensing data / measurements / coarse / fine-grain layer information (e.g., PHY / MAC / data link layer) / digital gain / RF filters / front-end switches / DC offset / correction / IQ-compensation settings / environmental effects on radio signal propagation / channel input to output transformation / environment Stable behavior of environment profile / state profile / wireless channel measurements / Received Signal Strength Indicator (RSSI) / Channel State Information (CSI) / Channel Impulse Response (CIR) / Channel Frequency Response (CFR) / Channel Response (CR) / Frequency component characteristics (e.g., subcarriers) / Channel characteristics / Channel filter response, auxiliary information, data / meta / user / account / access / security / session / status / monitoring / device / network / household / neighborhood / environment / real-time / sensor / storage / encrypted / compressed / protected data, identity / identifier / identification information.

[0083] Each CI may be associated with a timestamp / time of arrival / frequency band / signature / phase / amplitude / trend / characteristics, frequency-like characteristics, time / frequency / time-frequency domain elements, orthogonal / non-orthogonal decomposition characteristics of the signal passing through the channel. The timestamps of the TSCIs may be irregular and may be corrected (e.g., by interpolation / resampling) to be regular, at least over a sliding time window.

[0084] A TSCI may be / include link-wise TSCIs associated with antennas of a Type 1 device and antennas of a Type 2 device. For a Type 1 device with M antennas and a Type 2 device with N antennas, there may be MN link-wise TSCIs.

[0085] The CI / TSCI may be pre-processed / processed / post-processed / stored / retrieved / sent / received. Some modem / radio condition parameters may be held constant. The modem parameters may be applied to the radio subsystem and may represent radio conditions. The motion detection signal (e.g., a baseband signal, then decoded / demodulated packets) may be obtained by processing (e.g., downconverting) a radio signal (e.g., an RF / WiFi / LTE / 5G / 6G signal) by the radio subsystem using the radio conditions represented by the stored modem parameters. The modem parameters / radio conditions may be updated (e.g., using previous modem parameters / radio conditions). Both the previous / updated modem parameters / radio conditions may be applied in the radio subsystem (e.g., to process the signal / decode the data). In the disclosed system, both may be obtained / compared / analyzed / processed / monitored.

[0086] Each CI may include N1 CI components (CICs) (e.g., time / frequency domain components, decomposed components), each with a corresponding CIC index. Each CIC may include a real / imaginary / complex quantity, magnitude / phase / Boolean / flag, and / or some combination / subset. Each CI may include a vector / matrix / set / collection of CICs. The CICs of TSCIs associated with a particular CIC index may form a CIC time series. The TSCI may be divided into N1 time series of CICs (TSCICs), each associated with a respective CIC index. Characteristics / STI / MI may be monitored based on the TSCICs. Some TSCICs may be selected for further processing based on some criteria / cost function / signal quality metric (e.g., SNR, interference level).

[0087] Multi-component characteristics / STI / MI of multiple TSCICs (e.g., two components with indexes 6 and 7, or three components indexed at 6, 7, and 10) can be calculated. In particular, a k-component characteristic can be a function of k TSCICs with k corresponding CIC indices. When k=1, it is a single-component characteristic that can constitute / form a one-dimensional (1D) function because the CIC index spans all possible values. When k=2, a two-component characteristic can constitute / form a 2D function. In special cases, it can depend only on the difference between the two indices. In this case, it may constitute a 1D function. A sum characteristic can be calculated based on one or more multi-component characteristics (e.g., weighted average / aggregation). An object / motion / expression characteristic / STI / MI can be monitored based on any multi-component characteristic / total characteristic.

[0088] Characteristics / STI / MI may include: instantaneous / short-term / repeating / short-term / repeating / periodic / repeating / periodic / repeating / periodic / repeating / periodic / repeating / temporal / chronological / orthogonal / transformed / deterministic / probabilistic / primary / characteristic / significant / indicative / typical / prototypical / sustained / anomalous / sudden / abnormal / atypical / hazard / alarm / evolving / transient / transient quantity / characteristic / feature / information, cause and effect indicators / autocorrelation / covariance, autocorrelation function (ACF), spectrum / spectrography system / power spectral density, time / frequency function / transform / projection, initial / final / time / change / trend / trend / behavior / activity / history / profile / event, place / location / localization / spatial coordinates / change on map / route / navigation / tracking, line / rotation / horizontal / vertical / position / distance / displacement / height / velocity / acceleration / change / angular velocity, direction / orientation, size / length / width / height / orientation / area / volume / capacity, deformation / transformation, object / direction of motion / angle / shape / form / reduction / Magnification, behavior / activity / movement, occurrence, fall / accident / security / event, period / frequency / rate / cycle / rhythm / number / amount, timing / duration / interval, start / start / end / present / past / next time / amount / information, type / grouping / classification / composition, presence / absence / proximity / approach / retreat / entrance / exit, identity / identifier, head / mouth / eyes / breathing / heart / hands / arms / body / gestures / legs / gait / organ characteristics, tidal volume / breathing depth / air velocity / inspiration / exhalation time / ratio, walking / tools / machines / complex movements, signals / movement characteristics nature / information / features / statistics / parameters / magnitude / phase / degree / dynamics / anomalies / variations / detection / estimation / recognition / identification / display, function slope / derivatives / higher order derivatives / features / mapping / transformation of other properties, discrepancy / distance / similarity score / cost / metric, Euclidean / statistical / weighted distance, L1 / L2 / Lk norm, dot product / cross product, tag, quantity inspected, quantity consumed / unconsumed, state / physical / health / well-being / emotional / mental state, output response, any configuration / combination, and / or any related property / information / combination.

[0089] Test quantities may be calculated. Characteristics / STI / MI may be calculated / monitored based on CI / TSCI / features / similarity scores / test quantities. Static (or dynamic) segments / profiles may be analyzed to analyze CI / TSCI / features / features / functions / test quantities / features / STI / MI (e.g., target motion / motion presence / detection / estimation / recognition / identification information) to calculate / analyze / obtain / acquire / mark / present / display / highlight / store / communicate identification information. Test quantities may be based on CI / TSCI / features / feature functions / features / STI / MI. Test quantities may be processed / examined / analyzed / compared.

[0090] Test quantities consist of any / all functions of the following: data / vector / matrix / structure, characteristic / STI / MI, CI information (CII, e.g., CI / CIC / feature / magnitude / phase), directional information (DI, e.g., directional CII), dominant / representative / characteristic / indicative / key / typical / exemplary / salient / common / shared / typical / prototypical / average / regular / consistent / normal / non-normal / non-representative data / vector / matrix / structure, similarity / discrepancy / distance score / cost / metric, auto / cross-correlation / covariance, sum / average / weighted / trim / arithmetic mean / geometric mean / harmonic mean, variance / deviation / absolute / squared variance / mean / median / sum / standard deviation / differential / slope / variation / sum / absolute / squared variance / spread / min Dispersion / dispersibility, divergence / skewness / kurtosis / range / interquartile range / coefficient of variation / variance / L denominator / interquartile coefficient of variation / mean absolute value / squared difference / Gini coefficient / relative mean difference / entropy / maximum (max) / minimum (min) / median / percentile / quartile, variance-to-mean ratio, max-to-min ratio, variability / regularity / similarity measure, transient events / behavior, statistics / mode / likelihood / histogram / probability distribution function (pdf) / moment generating function / expectation function / value, behavior, repetition / periodicity / pseudoperiodicity, impulsivity / suddenness / occurrence / reproducibility, temporal profile / characteristics, time / minimum time / duration / period / frequency / trend / history, start / start / end time / quantity / number of times, movement classification / type, change, temporal / frequency / periodic change, etc.

[0091] The identification information / identity / identification information / ID may include a MAC address / ASID / USID / AID / UID / UUID, label / tag / index, web link / address, code / English code ID, name / password / account / account ID, and / or another ID (e.g., via software / firmware / user / hardware, hardwired, dongle). The ID may be stored / retrieved (e.g., stored locally / remotely / persistently / temporarily in a database / memory / cloud / edge / local / hub server). The ID may be associated with any of the following: user / customer / household / information / data / address / phone number / social security number, user / customer number / record / account, timestamp / duration / timing. The ID may be made available to Type 1 / Type 2 devices / sensing / SBP initiator / responder. ID is for registration / initialization / communication / identification / verification / detection / recognition / authentication / access control / cloud access / networking / social networking / logging / recording / cataloging / classification / tagging / associating / pairing / transactions / electronic transactions / intellectual property control (local / cloud / server / hub, type 1 / type 2 / neighborhood / user / other devices, user, etc.).

[0092] Objects include people / pets / animals / plants / users, baby / man / experts / staff / personnel / personnel / personnel / nurses / technicians / servicemen / patients / students / travellers / inmates / tracked objects, vehicles / cars / drones / robots / wagons / transporting machinery / movable objects / items / goods / parts / machines / lifts / elevators, goods / luggage / people / items / packaging / luggage / equipment / workflow / assembly line / warehouse / factory / stores / supermarkets / distribution / logistics / transportation / manufacturing / retail / wholesale / business centres / facility / cleaning tools in hubs, phones / computers / laptops / tablets / dongles / plug-ins / companions / tools / peripherals / accessories devices, wearables, furniture, appliances, amenities, gadgets, IoT, network, smart, portable devices, watches, glasses, speakers, toys, strollers, keys, wallets, handbags, backpacks, goods, cargo, luggage, equipment, motors, machines, appliances, tables, chairs, air conditioners, doors, windows, heaters, fans, lights, fixtures, stationary objects, televisions, cameras, audio, video, surveillance equipment, parts, tickets, parking passes, passes, airline tickets, credit cards, plastic cards, access cards, fixed, mutable, intangible objects, mass, solids, liquids, gases, fluids, smoke, fire, flames, signs, electromagnetic (EM) sources, media, and / or another object.

[0093] An object may have multiple parts, each with different motion (e.g., change of position / location / orientation). The object may be a person walking forward. While walking, his left / right hands may move in different directions with different instantaneous motions / velocities / accelerations.

[0094] The object may / may not be communicatively coupled to several networks such as WiFi, MiFi, 4G / LTE / 5G / 6G / 7G / 8G, Bluetooth / NFC / BLE / WiMax / Zigbee / mesh / adhoc networks, etc. AC-powered machinery that moves during installation, cleaning, maintenance, and renewal is bulky. It may be placed on / in a moving platform such as an elevator / conveyor / lift / pad / belt / robot / drone / forklift / car / boat / vehicle, etc. Type 1 / Type 2 devices can be attached to / moved on the object. The Type 1 / Type 2 device may be part of a portable / another device with a module (e.g., a module / device with a module, large / large / small / large / large / lightweight, e.g., coin-sized / cigarette-pack-sized), and the Type 1 / Type 2 / portable / another device may be attached to / not moved with the object and may have a wireless (e.g., via Bluetooth / BLE / Zigbee / NFC / WiFi) or wired (e.g., USB / micro USB / Firewire / HDMI) connection (e.g., via WiFi / cellular network) with a nearby device for network access. The nearby device may be an object / phone / AP / IoT / device / appliance / peripheral / amenity / furniture / vehicle / gadget / wearable / network / computing device. The nearby device may be connected to some server (e.g., a cloud server via a network / Internet). It may be portable / mobile or not, and may or may not move with the object. Type 1 / Type 2 / Portable / Proximity / Other devices can be battery / solar / DC / AC / other powered, swappable / non-swappable, rechargeable / non-rechargeable, or wirelessly charged.

[0095] Type 1 / Type 2 / Portable / Proximity / Other devices may include any of the following: Computer / Laptop / Tablet / Pad / Phone / Printer / Monitor / Battery / Antenna, Peripherals / Accessories / Sockets / Plugs / Chargers / Switches / Adapters / Dongles, Internet of Things (IoT), TV / Soundbar / HiFi / Speaker / Set-top Box / Remote Control / Panel / Gaming Console, AP / Cable / Broadband / Router / Repeater / Extender, Appliance / Utility / Fan / Refrigerator / Washer / Dryer / Microwave / Oven / Stove / Range / Light / Lamp / Pipe / Faucet / Lighting / Air Conditioner / Heater / Smoke Detector, Wearable Items / Watches / Eyeglasses / Goggles / Buttons / Bracelets / Chains / Jewelry / Rings / Belts / Clothing / Fabric / Shirts / Pants / Dresses / Gloves / Handwear / Shoes / Footwear / Hats / Headwear / Bags / Wallets / Purses / Makeup / Cosmetics / Decorative Items / Books / Magazines / Paper / Stationery / Signs / Posters / Displays / Printed Materials, Furniture / Fixtures / Table / Desk / Chair / Sofa / Bed / Cabinet / Shelf / Rack / Storage / Box / Bucket / Basket / Packaging / Car / Tile / Plate / Brick / Block / Mat / Panel / Curtain / Cushion / Pad / Carpet / Material / Building Material / Glass, Amenities / Sensor / Clock / Pot / Dishes / Container / Bottle / Can / Cookware / Plate / Cup / Bowl / Toy / Ball / Tool / Pen / Racket / Key / Bell / Camera / Microphone / Painting / Frame / Mirror / Coffee Maker / Door / Window, Food / Pill / Medicine, Implantable / Implantable / Gadget / Equipment / Apparatus / Instrument / Machine / Controller / Machine Tool, Garage Opener, Key / Plastic / Payment / Credit Card / Ticket, Solar Panel, Key Tracker, Fire Extinguisher, Trash Can / Dust Bin, WiFi Enabled Device, Smart Device / Machine / Machinery / System / House / Office / Building / Warehouse / Facility / Vehicle / Automobile / Bicycle / Motorcycle / Boat / Watercraft / Airplane / Cart / Wagon, House / Vehicle / Office / Factory / Building / Manufacturing / Production / Computing / Security / Other Devices.

[0096] One / two / more Type 1 / Type 2 / portable / proximity / another device / server may determine the initial characteristics / STI / MI of the object and / or share intermediate information. One of the Type 1 / Type 2 devices may move with the object (e.g., "Tracker Bot"). The other Type 1 / Type 2 device may not move with the object (e.g., "Origin Satellite", "Origin Register"). Both may have known characteristics / STI / MI. The initial characteristics / STI / MI may be calculated based on the known characteristics / STI / MI.

[0097] A venue can be a space such as: sensing area, room / house / home / office / workplace / building / facility / warehouse / factory / store / vehicle / real estate, indoor / outdoor / closed / semi-closed / open / semi-open / closed / air / floating / underground space / area / structure / enclosure, wood / glass / metal / material / structure / frame / beam / panel / column / wall / floor / door / ceiling / window / cavity / gap / opening / reflective / refractive medium / fluid / structural material / space / region with fixed / adjustable layout / shape, human body / animal / plant body / cavity / organ / bone / blood / blood vessel / air pipe / wind pipe / tooth / soft / hard / non-hard tissue, manufacturing / repair / maintenance / mining / parking / storage / transportation / ship / logistics / sports / recreation / amusement / public / entertainment / government / community / elderly / seniors / space facility / terminal / hub, logistics center / store, machinery / engine / equipment / assembly line / workflow, city / rural / suburban / metropolitan area, stairs / escalator / elevator / hallway / passage / tunnel / cave / cavern / waterway / duct / pipe / tube / lift / well / passage / roof / basement / sinkhole / alley / road / path / highway / sewer / ventilation system / network, car / truck / bus / van / container / ship / boat / submersible / train / tram / plane / mobile home, stadium / city / playground / park / field / track / court / gym / hall / market / supermarket / square / plaza / construction site / hotel / museum / school / hospital / university / garage / mall / airport / train station / bus stop / terminal / hub / platform, valley / forest / tree / terrain / landscape / garden / park / patio / land, and / or gas / oil / water pipe / line. A venue may consist of the interior or exterior of a building / facility. A building / facility may be one-story or multi-story, with some underground.

[0098] Events can be monitored based on the TSCI. Events can be object / motion / gesture / walk-related, such as falls, rotations / hesitations / poses, impacts, shocks (e.g., hitting a punching bag / door / bed / chair / table / desk / cabinet / box / another person / animal / bird / flying / ball / bowling / tennis / football / volleyball / soccer / baseball / basketball), two-body actions (e.g., balloon liberator / catching fish / moldable clay / paper / typing computer), movement in a garage, smartphone / person walking in a venue, autonomous / mobile object / machine motion (e.g., vacuum cleaner / utility / self-driving vehicle / car / vehicle / drone), etc.

[0099] Tasks may include: (a) Sensing task, i.e., monitoring / sensing / detection / recognition / estimation / verification / identification / authentication / classification / location / guidance / navigation / tracking / counting of / in any of the following: objects / vehicles / machines / tools / humans / babies / elderly / patients / intruders / pet presence / proximity / activity / daytime activity / health status / breathing / vital signs / heart rate / health status / sleep / sleep stages / gait / location / distance / speed / acceleration / navigation / tracking / motion / safety / hazard / fall / intrusion / security / life threat / movement / motion / decline / pattern / cyclical / repeating / periodic / steady / regular / transient / sudden / suspicious movement / irregularity / trend / change / breathing / human biometric information / environmental information / gate / gesture / room / area / zone / street; (b) Computational tasks, i.e., any of the following: signal processing / pre-processing / post-processing / conditioning / noising / calibration / analysis / feature extraction / transformation / mapping / supervised / unsupervised / semi-supervised / discrimination / machine / deep learning / learning / clustering / learning / PCA / eigendecomposition / frequency / time / function decomposition / neural network / map-based / model-based processing / correction / shape estimation / analysis computation; (c) IoT tasks, i.e., any of the following: smart tasks for venues / users / objects / humans / pets / houses / homes / offices / workplaces / buildings / facilities / warehouses / factories / stores / vehicles / properties / structures / assembly lines / IoT / devices / systems, energy / power management / transfer, wireless power transfer, interaction / collaboration with users / objects / intruders / humans / animals (presence / motion / gesture / walk / activity / behavior / voice / command / instruction / query / music / sound / image / video / location / movement / hazard / threat detection / recognition / monitoring / analysis / response / execution / synthesis, dialogue generation / retrieval / playback / display / rendering / composition / exchange / response / presentation / experience / media / multimedia / representation / sound / voice / music / image / video / animation / web page) page / text / message / notification / attention / inquiry / warning, user / intruder / object input / motion / gesture / location / activity detection / recognition / monitoring / interpretation / analysis / recording / storage), activation / control / setting (on / off / control / lock / unlock / open / close / adjust / setting) of devices / systems (e.g., vehicles / drones / electrical / machines / air conditioning / heating / lighting / ventilation / learning / entertainment / IoT / security / siren / access systems / devices / doors / windows / garages / lifts / elevators / escalators / speakers / TVs / lighting / peripherals / accessories / wearables / furniture / home appliances / amenities / gadgets / alarms / cameras / games / coffee / cooking / heaters / fans / housekeeping / household / office machines / devices / robots / vacuum cleaners / assembly lines), (d) Various tasks, i.e., any of the following: transmitting data / parameters / analysis / derived data / encoding / encrypting / storing / analyzing, upgrading / managing / configuring / adjusting / broadcasting / synchronizing / networking / encrypting / communicating / protecting / compressing / storing / database / archiving / querying / cloud computing / presentation / augmented / virtual reality / other processing / tasks. Tasks may be performed by some of the Type 1 / Type 2 / nearby / portable / other devices and / or hub / local / edge / cloud servers.

[0100] Tasks may also include: detect / recognize / monitor / locate / interpret / analyze / record / store user / visitor / intruder / object / pet, interact / engage / talk / converse / exchange with user / object / visitor / intruder / human / baby / pet, detect / locate / recognize / monitor / analyze / interpret / learn / train / respond / perform / synthesize / generate / record / store / summarize health / health status / daily life / activities / behaviors / patterns / exercise / diet Intake / Toilet / Work / Play / Rest / Sleep / Relax / Danger / Routine / Limit / Habit / Tendency / Normal / Abnormal / Regularity / Irregularity / Change / Presence / Action / Gesture / Walk / Facial Expression / Emotion / State / Voice / Command / Instruction / Question / Inquiry / Music / Sound / Location / Movement / Fall / Threat / Discomfort / Illness / Environment / , Generate / Search / Play / Display / Render / Dialogue Synthesis / Exchange / Response / Presentation / Report / Experience / Media / Multimedia / Representation / Sound / Voice / Music / Image / Video / Animation / Web Page / Text / Message / Notification / Reminder / Inquiry / Warning, User / Intruder / Object Input / Movement / Gesture / Location / Activity Detection / Recognize / Monitor / Interpret / Analyze / Record / Store), Detect / Check / Monitor / Location / Manage / Control / Adjust / Configure / Lock / Unlock / Alarm / Disarm / Open / Close / Whole / Partial / Activate / Activate Turning on / off (e.g., vehicle / robot / drone / electrical / machine / air conditioning / heating / ventilation / HVAC / lighting / cleaning / entertainment / IoT / security / siren / access system / device / item / component, door / window / garage / lift / elevator / escalator / speaker / TV / lighting / peripherals / accessories / wearable / furniture / appliance / amenity / gadget / alarm / camera / game / coffee / cooking / heater / fan / housekeeping / home / office machine / device / vacuum cleaner / assembly line / window / garage / door / blind / curtain / panel / solar panel / sunshade), detecting / monitoring / locating where a user / pet is doing something (e.g., sitting on the couch / sleeping / sleeping in the bedroom / running on a treadmill / cooking / watching TV / eating in the kitchen / dining room / going up / down stairs / outside / inside / using the toilet),Do something (e.g. do something automatically upon detection (generate a message / response / alert / clarification / notification / report), do something automatically for the user upon detection of their presence, turn on / off / alarm / control / adjust / dimming lights / music / radio / TV / HiFi / STB / computer / speaker / smart device / air conditioning / ventilation / heating system / curtains / light shades, turn on / off / preheat / control a coffee maker / kettle / cooker / oven / microwave / another cooking appliance, check / manage temperature / settings / weather forecast / phone call / message / email / system check, present / interact / engage / dialogue / conversation (via smart speaker / display / screen, via web page / email / messaging system / notification system etc.)

[0101] When a user arrives home in their car, the tasks may be to automatically detect the user / car's approach, open the garage door upon detection, turn on the driveway / garage light as the user approaches the garage, and / or turn on the air conditioner / heater / fan. When a user enters the house, automatically turn on the porch light / garage light, play a greeting message with the user's favorite music / radio / news / channel, monitor the user's mood, adjust the lighting / sound environment depending on the mood / current / impending event (e.g., romantic lighting / music because the user is having dinner with his / her girlfriend soon), microwave a hot meal prepared in the morning, check the weather forecast for tomorrow / news, user's interests, check the calendar / to-do list, answer the phone / messaging system / check email, verbally report using a dialogue system / speech synthesis, and / or play TV / entertainment. Using audio tools, using visual tools such as entertainment systems / computers / notebooks / displays / lights / colors / brightness / pattern symbols, haptics / virtual reality / gestures / tools, using smart devices / appliances / materials / furniture / fixtures, using servers / hub devices / cloud / fog / edge servers / home / mesh networks, using messaging / notification / communication / scheduling / email tools, using UI / GUI, using scents / smells / fragrances / tastes, using neural / nervous systems / tools, or any combination) to prepare a user for someone's birthday / phone call. A task might be to preemptively turn on an air conditioning / heating / ventilation system and / or preemptively adjust the temperature setting on a smart thermostat. When a user moves from the entrance to the living room, the task might be to turn on the living room lights, open the living room curtains, open the windows, turn off the entrance light behind the user, turn on the TV / set-top box, set the TV to the user's preferred channel, and / or adjust appliances according to the user's preferences / conditions / states (e.g., adjust the lighting, select / play music to create a romantic atmosphere).

[0102] If a user wakes up in the morning, the tasks might be to detect the user's motion in the bedroom, open the blinds / curtains / window, turn off the alarm clock, adjust the night-to-day temperature profile, turn on the bedroom light, turn on the toilet light as the user approaches the toilet, check the radio / streaming channel, play the morning news, turn on the coffee machine, turn on the preheated water, and / or turn off the security system. If a user walks from the bedroom to the kitchen, the tasks might be to turn on the kitchen / hallway light, turn off the bedroom / toilet light, move music / messages / reminders from bedroom to bedroom, turn on the kitchen TV, change the TV to the morning news channel, lower the kitchen blinds, open the kitchen window, unlock the back door so the user can check the back yard, and / or adjust the kitchen temperature setting.

[0103] When the user leaves home for work, the task is to detect when the user has left the house, play a farewell / welcome message, open / close the garage door, turn on / off the garage / driveway lights, close / lock all windows / doors (in case the user has forgotten), turn off appliances (stove / microwave / oven, etc.), turn on / arm the security system, adjust the lighting / AC / heating / ventilation systems to an "away" profile to save energy, and / or send alerts / reports / updates to the user's smartphone.

[0104] Motion may include: no motion, motion sequence, stationary / non-moving motion, movement / position / location change, daily / weekly / monthly / yearly / repeated / activity / behavior / routine, transient / time varying / fall / repeated / periodic / quasi-periodic motion / breathing / heartbeat, deterministic / non-deterministic / stochastic / chaotic / random motion, complex / complex motion, non / quasi / cyclo / stationary random motion, change in electromagnetic properties, human / animal / plant / body / machine / vehicle / drone motion, Air / wind / weather / water / fluid / ground / earthquake motion, human-machine interaction, normal / abnormal / danger / warning / suspicious motion, impending / rain / fire / flood / tsunami / explosion / collision, head / face / eyes / mouth / tongue / neck / fingers / hands / arms / shoulders / upper / lower / body / chest / abdomen / waist / legs / feet / joints / knees / elbows / skin / subcutaneous / subcutaneous tissue / blood vessels / organs / heart / lungs / stomach / intestines / bowels / eating / breathing / speaking / singing / dancing / coordination motion, facial / eye / mouth expressions, and / or hands / arms / gestures / walking / UI / keystrokes / typing strokes.

[0105] A Type 1 / Type 2 device may include a heterogeneous IC, a low-noise amplifier (LNA), a power amplifier, a transmit / receive switch, a media access controller, a baseband radio, and / or a 2.4 / 3.65 / 4.9 / 5 / 6 / sub-7 / over-7 / 28 / 60 / 76 GHz / other radio. The heterogeneous IC may include a processor, memory, software, firmware, and instructions. It may support broadband / wireless / mobile / mesh / cellular networks, WLAN / WAN / MAN, standards / IEEE / 3GPP / WiFi / 4G / LTE / 5G / 6G / 7G / 8G, IEEE 802.11 / a / b / g / n / ac / ad / af / ah / ax / ay / az / be / bf / 15 / 16, and / or Bluetooth / BLE / NFC / Zigbee / WiMax.

[0106] The processor may include any of general-purpose / special-purpose / embedded / multi-core processors, microprocessors / microcontrollers, multi / parallel / CISC / RISC processors, CPUs / GPUs / DSPs / ASICs / FPGAs, and / or logic circuits. The memory may include non-volatile, RAM / ROM / EPROM / EEPROM, hard disks / SSDs, flash memory, CD- / DVD-ROMs, magnetic / optical / organic / storage systems / networks, network / cloud / edge / local / external / internal storage, and / or any non-transitory storage medium. The set of instructions may comprise machine-executable code in hardware / ICs / software / firmware and may be embedded / preloaded / loaded at boot-up / on-the-fly / on-demand / pre-installed / installed / downloaded.

[0107] Processing / pre-processing / post-processing may be applied to data (e.g., TSCI / features / characteristics / STI / MI / test quantities / intermediate / data / analysis) and may have multiple steps. Steps / processing / pre-processing / post-processing may include any of the following: operand / LOS / non-LOS / single-link / multi-link / component / item / quantity arithmetic functions, magnitude / norm / phase / feature / energy / time axis / similarity / distance / characteristic score / measure calculation / extraction / correction / cleaning, linear / non-linear / FIR / IIR / MA / AR / ARMA / Kalman / particle filtering, low-pass / band-pass / high-pass / median / rank / quartile / percentile / mode / selection / adaptive filtering, interpolation / extrapolation / decimation / subsampling / upsampling / resampling,Matched filtering / enhancement / restoration / noising / smoothing / conditioning / spectral analysis / mean subtraction / removal, linear / nonlinear / inverse / frequency / time transform, Fourier transform (FT) / DTFT / DFT / FFT / wavelet / Laplace / Hilbert / Hadamard / trigonometric / sine / cosine / DCT / power of two / sparse / fast / frequency transform, zero / cyclic / padding, graph-based transformation / processing, decomposition / orthogonal / non-orthogonal / complete projection / eigendecomposition / SVD / PCA / ICA / compressed sensing, grouping / folding / sorting / comparison / soft / hard / threshold / clipping, first / second / higher order differentiation / integration / convolution / multiplication / division / addition / subtraction, local / global / maximization / minimization, recursion / iteration / constraint / batch processing, least mean square / absolute error / deviation, cost function optimization, neural network / detection / recognition / Classification / identification / estimation / labeling / association / tagging / mapping / remapping / learning / clustering / machine / supervised / unsupervised / semi-supervised learning / network, vector / quantization / encryption / compression / matching pursuit / scrambling / coding / storing / retrieval / send / receive / time domain / frequency domain / normalization / scaling / representation / combining / segmentation / tracking / monitoring / shape / silhouette / motion / activity / analysis, pdf / histogram estimation / importance / Monte Carlo sampling, error detection / protection / correction, do nothing, time-varying / adaptive processing, conditioning / weighting / averaging / selected components / over links, arithmetic / geometric / harmonic / trimmed mean / centroid / medoid calculation, morphological / logical operations / permutation / combination / sorting / AND / OR / XOR / sum / intersection, vector operations / addition / subtraction / multiplication / division, and / or other operations. Operations may be applied individually or jointly. Acceleration using GPU / DSP / coprocessor / multi-core / multi-processing can also be applied.

[0108] Functions may include: property / characteristic / magnitude / phase / energy, scalar / vector / discrete / continuous / polynomial / exponential / logarithmic / trigonometric / transcendental / logical / piecewise / linear / algebraic / nonlinear / circular / piecewise linear / real / complex / vector-valued / reciprocal / absolute value / index / limit / floor / circular / sign / composite / slide / shift function, differentiation / integration, function of functions, one-to-one / one-to-many / many-to-one / many-to-many functions, mean / mode / median / percentile / max / min / range / statistics / histogram, local / global max / min / zero crossing, variance / variation / spread / variance / deviation / standard deviation / divergence / range / interquartile range / total variation / absolute value / total deviation, arithmetic mean / geometric mean / harmonic mean / trimmed mean / squared / cubed / root / power, thresholding / clipping / rounding / truncation / quantization / approximation, time functions processed by operations (e.g. filtering), sine / cosine / tangent / bitangent / elliptic / parabolic / hyperbolic / game / zeta function, stochastic / stochastic / random / ergodic / stationary / deterministic / periodic / repeated functions, inverse transform / frequency / discrete time / Laplace / Hilbert / sine / cosine / trigonometric / wavelet / integer / power of two / sparse transform, Orthogonal / non-orthogonal / eigenprojection / decomposition / eigenvalues / singular values / PCA / ICA / SVD / compressed sensing / neural network / feature extraction / function of moving window of nearby items in time series / filtering function / convolution / short-time / discrete transform / Fourier / cosine / sine / Hadamard / wavelet / sparse transform / matching pursuit / approximation, graph-based processing / transform / graph signal processing, classification / identification / class / group / category / labeling, processing / preprocessing / postprocessing, machine / learning / detection / estimation / feature extraction / learning network / feature extraction / noise removal / signal enhancement / Coding / Encryption / Mapping / Vector Quantization / Remapping / Low Pass / High Pass / Band Pass / Matched / Kalman / Particle / FIR / IIR / MA / AR / ARMA / Median / Mode / Adaptive Filtering, 1st / 2nd / Higher Order Differentiation / Integration / Zero Crossing / Smoothing, Up / Down / Random / Importance / Monte Carlo Sampling / Resampling / Transformation, Interpolation / Extrapolation, Short Term / Long Term Statistics / Auto / Cross Correlation / Moment Generating Function / Time Average / Weighted Average, Special / Bessel / Beta / Gamma / Gaussian / Poisson / Integral Complementary Error Functions.

[0109] The sliding time window may vary in width / size over time. It may be initially small or large to allow for fast and accurate imaging, and over time, it increases / decreases to a steady-state size comparable to the frequency / period / transient motion duration / characteristics / STI / MI of the monitored motion. The window size / time shift between adjacent windows may be constant / adaptive / dynamic / automatically / changed / adjusted / fluctuated / modified (e.g., based on battery life / power consumption / available computing power / changes in the amount of monitored object / nature of monitored motion / user request / selection / instruction / command).

[0110] The characteristic / STI / MI may be determined based on characteristic values / points of the function and / or the function's associated arguments (time / frequency, etc.). The function may be the result of a regression. The characteristic values / points may include the local / global / constrained / significant / first / second / i-th maximum / minimum / extremum / zero-crossing (e.g., with positive / negative time / frequency / argument) of the function. A local signal-to-noise ratio (SNR) or SNR-like parameter may be calculated for each pair of adjacent local maxima (peaks) / local minima (valleys) of the function, which may be some function (e.g., linear / logarithmic / exponential / monotonic / power / polynomial) of the fraction or difference of the quantity (e.g., power / magnitude) of the local maximum relative to the quantity of the local minimum. A local maximum (or minimum) may be significant if its SNR is greater than a threshold and / or if its amplitude is greater (or less) than another threshold. The local maximum / minimum may be selected / identified / calculated using a persistence-based approach. Several significant local maxima / minima may be selected based on selection criteria (e.g., quality criteria / conditions, strongest / consistently significant peak within a range). Unselected significant peaks may be stored / monitored as "reserved" peaks for use in future selections in future sliding time windows. For example, a particular peak (e.g., at a particular argument / time / frequency) may appear consistently over time. Initially, it may be significant but not selected (because other peaks may be stronger). Later, it may become stronger / consistently dominant. Once selected, it may be backtraced in time and selected at an earlier time to replace the previously selected peak (momentarily strong / dominant, but not persistent / consistent). Peak consistency may be measured by the trace or duration of significance. Alternatively, local maxima / minima may be selected based on a finite state machine (FSM). The decision threshold is time-varying and may be adjusted adaptively / dynamically (e.g., based on backtrace timing / FSM, or data distribution / statistics).

[0111] A similarity score (SS) / component SS can be computed based on two temporally adjacent CIs / CICs, one TSCI, or two different TSCIs. Pairs may originate from the same / different (one or more) sliding windows. SS or component SS may include time reversal resonating strength (TRRS), auto / cross-correlation / covariance, dot product of two vectors, L1 / L2 / Lk / Euclidean / statistical / weighted / distance score / norm / metric / quality index, signal quality criteria, statistical features, discriminative scores, neural networks / deep learning networks / machine learning / training / discrimination / weighted average / preprocessing / noising / signal conditioning / filtering / time correction / timing compensation / phase offset compensation / transform / component-wise operation / feature extraction / FSM, and / or other scores.

[0112] Any threshold may be fixed (e.g., 0, 0.5, 1, 1.5, 2), predetermined, and / or adaptively / dynamically determined (e.g., by an FSM or based on time / space / location / antenna / path / link / condition / battery life / remaining battery life / available resources / power / computing power / network bandwidth). A threshold may be applied to a test quantity to distinguish between two events / conditions / situations / states. Data (e.g., CI / TSCI / feature / similarity score / test quantity / feature / STI / MI) may be collected under A / B in a training situation. Test quantities (e.g., their distributions) calculated based on the data may be compared under A / B to select a threshold based on several criteria (e.g., maximum likelihood (ML), maximum a posteriori probability (MAP), discriminative training, minimum type 1 (or 2) error for a given type 2 (or 1) error, quality criteria, signal quality conditions). The threshold may be adjusted (e.g., to achieve different sensitivities) based on (e.g., object / movement / direction / action / characteristic / STI / MI / size / property / habit / behavior / venue / feature / fixture / furniture / barrier / material / living thing / thing / boundary / surface / location / map / machine / model / event / state / situation / condition / time / timing / duration / status / history / user / preference). The iterative algorithm may stop after N iterations, after a timeout period, or after meeting a condition (e.g., update amount greater than threshold) where the test quantity may be fixed / adaptively / dynamically adjusted.

[0113] Searching for local extrema, constrained / minimization / maximization, statistical / dual / constrained / convex / global / local / combinatorial / infinite-dimensional / multi-objective / multimodal / non-differential / particle swarm / simulation-based optimization, linear / nonlinear / quadratic / higher-order regression, linear / nonlinear / stochastic / constrained / mechanical / mathematical / non-connected / convex / semi-defined / cone / interior / fractional / integer / sequential / quadratic programming, conjugate / gradient / subgradient / coordinate / contraction descent, Newton / complex / Iteration / point / elliptic / quasi-Newton / interpolation / memory / genetics / evolution / pattern / gravitational search / algorithms, constraint satisfaction, calculus of variations, optimal control, space mapping, heuristics / metaheuristics, numerical analysis, simultaneous perturbation stochastic approximation, stochastic tunneling, dynamic relaxation, hill climbing, simulated annealing, differential evolution, robust / line / tab / reaction search / optimization, curve fitting, least squares, calculus of variations, and / or variational methods. May be associated with objective functions / loss functions / cost functions / utility functions / fitness functions / energy functions.

[0114] Regression may be performed using a regression function to fit the data or a function of the data (e.g., ACF / transformed / mapped) within the regression window. The length / position of the regression window may be changed during the iterations. The regression function may be linear / quadratic / cubic / polynomial / other. The regression may minimize either the mean / weighted / absolute / squared deviation, error, aggregate / component / weighted / mean / absolute / squared / higher order / other error / cost (e.g., in a projected domain / selected axis / orthogonal axis), a robust error for smaller error magnitudes (e.g., a first error (e.g., squared), a second error (e.g., absolute) for larger error magnitudes), and / or a weighted sum / average of multiple errors (e.g., absolute / squared error). Errors associated with different links / paths may have different weights (e.g., links with less noise may have higher weights). The regression parameters (e.g., For example, the maximum / minimum regression error of the regression function in the regression window, the time offset relative to the window position / width) can be initialized and / or updated during the iteration (e.g., based on target values / ranges / profiles, characteristics / STI / MI / test quantities, object motion / amount / count / location / status, past / current trends, location / amount / distribution of extreme values ​​in the previous window, signal carrier / subcarrier frequency / bandwidth, amount of antenna associated with the channel, noise characteristics, histogram / distribution / center / F-distribution, and / or thresholds). Upon convergence, the current time offset becomes the center / left / right (or a fixed relative position) of the regression window.

[0115] In a presentation, information may be displayed / presented (e.g., using a venue map / environment model). Information may include: current / past / corrected / approximate / map / location / speed / acceleration / zone / region / area / segmentation / coverage area, direction / route / trace / history / traffic / summary, frequently visited areas, customer / crowd events / distribution / behavior, crowd control information, acceleration / speed / vital signs / breathing / heart rate / activity / emotion / sleep / status / rest information, motion statistics / MI / STI, presence or absence of moving body / person / pet / object / vital signs, gesture (e.g., hand / arm / leg / body / head / face / mouth / eye) / meaning / control (controlling a device using gestures), location-based gesture control / motion interpretation, ID / identifier (ID) (e.g., object / person / user / pet / zone / region, device / machine / vehicle / drone / car / boat / bicycle / TV / air conditioner / fan / self-guided machine / device ID / identifier). Objects / people / users / pets / zones / areas, devices / machines / vehicles / drones / cars / boats / bicycles / TVs / air conditioners / fans / self-guided machines / devices / vehicles), environment / weather information, gestures / gesture control / motion traces, earthquakes / explosions / storms / rain / fires / temperature, collisions / shocks / vibrations, events / doors / windows / opening / closing / falls / accidents / burning / freezing / water / wind / air movement events, recurring / quasi-periodic events (e.g., running on a treadmill, jumping, skipping rope, somersaults, etc.), and / or vehicle events. Location may be in one / two / three dimensions (e.g., represented / represented as 1D / 2D / 3D rectangular / polar coordinates), relative (e.g., with respect to a map / environment model), or relational (e.g., at / near / distance from / halfway between two points, around a corner, upstairs, on a tabletop, on the ceiling, on the floor, on a sofa).

[0116] Information (e.g., location) may be marked / displayed with some kind of symbol. The symbol may change over time / flash / pulse, changing color / intensity / size / orientation. The symbol may be a number reflecting an instantaneous quantity (e.g., analysis / gesture / state / status / action / motion / breathing / heart rate, temperature / network traffic / connectivity / remaining power). The symbol / size / orientation / color / intensity / rate of change / characteristics may reflect the respective motion. Information may be textual or presented visually / verbally (e.g., using pre-recorded voice / speech synthesis) / mechanically (e.g., animated gadget, moving parts).

[0117] User devices may include smartphones / tablets / speakers / cameras / displays / TVs / gadgets / vehicles / appliances / devices / IoT, devices with UI / GUI / audio / voice / recording / capture / sensor / playback / display / animation / VR / AR (augmented reality) / voice (assistance / recognition / synthesis) capabilities, and / or tablets / laptops / PCs.

[0118] The map / floor plan / environment model (e.g., home / office / building / store / warehouse / facility) may be 2 / 3 / higher dimensional. It may change / evolve over time (e.g., rotate / zoom / move / jump on the screen). Walls / windows / doors / entrances / exits / restricted areas may be marked. It may contain multiple layers (overlays). It may include maintenance maps / models including water pipes / gas pipes / cables / air ducts / crawl spaces / ceiling / underground layouts.

[0119] A venue may be divided / subdivided / zoned / grouped into multiple zones / regions / sectors / sections / territories / districts / districts / neighborhoods / areas / stretches / expanses such as bedroom / living / dining / resting / storage / utility / warehouse / meeting / work / hallway / kitchen / foyer / garage / ground floor / second floor / office / reception room / area / region etc. It may be presented in a map / floor plan / model with presentation properties (e.g. brightness / intensity / luminance / color / chrominance / texture / animation / flash / rate).

[0120] An example of the disclosed system / apparatus / method: Stephen and his family want to install the disclosed wireless motion detection system to detect motion in their 2,000-square-foot, two-story townhouse in Seattle, Washington. Because his house has two staircases, Stephen decides to use one Type 2 device (named A) and two Type 1 devices (named B and C) on the first floor. The first floor has three rooms lined up: the kitchen, dining room, and living room, with the dining room in the middle. A is in the dining room, B is in the kitchen, and C is in the living room, dividing the first floor into three zones (dining room, living room, kitchen). When motion is detected by the AB pair and / or AC pair, the system analyzes the TSCI / features / characteristics / STI / MI and associates the motion with one of the three zones.

[0121] When Stephen and his family go on holiday camping, he turns on the motion detection system using a mobile phone app (e.g., an Android phone app or an iPhone app). When the system detects motion, an alert signal (e.g., SMS, email, push message to the mobile phone app, etc.) is sent to Stephen. If Stephen pays a monthly fee (e.g., $10 / month), a service company (e.g., a security company) receives the alert signal via a wired (e.g., broadband) / wireless (e.g., WiFi / LTE / 5G) network and performs security procedures (e.g., calling Stephen to confirm the problem, sending someone to check at his home, contacting the police on Stephen's behalf).

[0122] Stephen loves his elderly mother and cares about her well-being when she is home alone. When his mother is at home during a family break (e.g., work / shopping / vacation), Stephen uses his mobile app to turn on the motion detection system to ensure his mother is ok. He uses the mobile app to monitor his mother's motion at home. When Stephen uses the mobile app to see his mother circling the house in three areas, according to her daily routine, Stephen knows that his mother is ok. Stephen is grateful that the motion detection system can help him monitor his mother's well-being while he is away from home.

[0123] On a typical day, Mom wakes up at 7 AM, spends 20 minutes cooking breakfast in the kitchen, and 30 minutes eating breakfast in the dining room. She then does her daily exercise in the living room, and then sits on the couch in the living room and watches her favorite TV show. The motion detection system allows Stephen to see the timing of motion in three areas of the house. If the motion agrees with the daily routine, Stephen knows that Mom should be generally fine. However, if the motion pattern is abnormal (e.g., no motion until 10 AM or too much time in the kitchen / stillness), Stephen suspects something is wrong and calls Mom to check on her. Stephen can even get someone (e.g., family member / neighbor / paid staff / friend / social worker / service provider) to check on Mom.

[0124] One day, Stephen feels like relocating the device. He simply unplugs it from its AC power plug and plugs it into another AC power plug. He's happy that the motion detection system is plug-and-play, and repositioning doesn't affect the system's operation. It works as soon as he powers it on.

[0125] Stephen then decides to install a similar setup (i.e., one Type 2 and two Type 1 devices) on the second floor to monitor the bedrooms on the second floor. Again, the system is very easy to set up; he simply plugs the Type 2 and Type 1 devices into AC power plugs on the second floor. No special installation is required. He can monitor motion on both the ground and second floors using the same mobile app. Each Type 2 device on the ground and second floors can interact with all Type 1 devices on the ground and second floors. Stephen has more than twice the capacity of the combined system.

[0126] The disclosed system can be applied to many applications. The Type 1 / Type 2 devices can be any WiFi-enabled device (e.g., smartphone IoT / IoT / appliance / STB / refrigerator / speaker / STB / refrigerator / fan / heater / fan / air conditioner / router / tablet / computer / tablet / plug / pipe / lamp / smoke detector / furniture / shelf / cabinet / door / lock / sofa / table / chair / piano / instrument / wearable / watch / tag / key / ticket / belt / wallet / pen / hat / necklace / implant / phone / glasses / glass panel / gaming device) located in a home / office / table, ceiling, floor, or wall. They can be placed in a conference room to count people. They can form a welfare monitoring system to monitor the daily activities of the elderly and detect any signs of symptoms (e.g., dementia, Alzheimer's disease). They can be used in an infant monitor to monitor an infant's vital signs (breathing). They can be placed in a bedroom to monitor sleep quality and detect any sleep apnea. They may be placed in cars to monitor the health of passengers and drivers and detect sleepy drivers or babies left in hot cars. They may be used in logistics to prevent human trafficking by monitoring people hidden in trucks / containers. They may be deployed by emergency services in disaster areas to search for victims trapped in rubbish. They may be deployed in security systems to detect intruders.

[0127] By digging into the underlying multipath propagation characteristics and finding a proximity feature based on the correlation of adjacent subcarriers, we can distinguish between nearby and distant motion. This feature has been proven to be sensitive to the distance between a moving human and a device. However, because it relies on dynamic reflections from human targets, the proximity feature is highly sensitive to the strength of the movement. For example, if a person moves from large movements, such as walking, to small movements, such as reading, after approaching a device, the value of the proximity feature drops, causing a proximity detector to fail to continuously detect the nearby person. Recognizing that humans typically approach and move away from devices by walking, we can consider gait information as an auxiliary factor to support human proximity detection. Gait can be estimated from speed, which cannot be easily estimated from WiFi signals. Speed ​​can be estimated from the autocorrelation function (ACF) of the CSI based on a statistical model. By utilizing the ACF spectrum and speed information, the present teachings disclose a gait score, which serves as an indicator of the presence or absence of walking.

[0128] The present teachings further combine proximity features with gait information to disclose a state machine that accumulates temporal information and triggers or de-triggers proximity detection only when a transition state, such as approach or departure, is detected. In this manner, proximity detection continues after initial detection, even if the detected human makes subtle movements, until a departure state is identified.

[0129] In some embodiments, the present teachings disclose a robust WiFi-based proximity detector that takes into account the monitoring of walking during proximity detection. This proximity detector does not simply rely on dynamic signal reflections, where the extracted features depend on the intensity of the movement. In some embodiments, a novel walking score is disclosed to accurately assess the presence of walking by leveraging velocity estimated from the autocorrelation function (ACF) of channel state information (CSI). By combining this walking score with proximity features, the disclosed method effectively distinguishes between different transition patterns, enabling more reliable proximity detection. Furthermore, to enhance the stability of the detection process, the system employs a state machine to extract temporal information to effectively distinguish between nearby and distant people, ensuring continuous proximity detection even during subtle movements (e.g., typing on a keyboard), and enhancing the stability of the detection process. Through extensive experiments in diverse environments, the disclosed system has been validated and is capable of achieving a high overall detection rate while maintaining a low false alarm rate across various scenarios. The disclosed proximity detection method and system are beneficial for applications in security systems, healthcare monitoring, and smart environments, and there are opportunities for future improvements and other real-world implementations.

[0130] In wireless communications, channel state information (CSI) or channel frequency response (CFR) describes the propagation of a WiFi signal from a transmitter (Tx) to a receiver (Rx). The CSI estimated on a subcarrier of frequency f at time t is expressed as: TIFF2025066624000002.tif18158Here, X(t,f) and Y(t,f) represent the transmitted signal at Tx and the received signal at Rx, respectively. Since the transmitted WiFi signal experiences multiple reflections during propagation in an indoor environment, the corresponding CSI can be written as an aggregation of radio propagations along multiple different paths: TIFF2025066624000003.tif14158Here α l is the multipath coefficient of the l-th independent multipath component, and τ lis the associated propagation delay. In practice, due to imperfect synchronization of commercial WiFi devices, random noise often occurs in the CSI phase, which is difficult to remove efficiently. Therefore, a more reliable CSI amplitude can be obtained, which can be measured via the power response G(t,f) as follows: TIFF2025066624000004.tif14158

[0131] 1 shows a diagram of an example system 100 for proximity detection in accordance with some embodiments of the present disclosure. As depicted in FIG. 1, system 100 includes a CSI preprocessing module 110, a proximity monitoring module 120, a walking detection module 130, and a state transition module 140. In some embodiments, system 100 is a robust proximity detector based on WiFi signals.

[0132] CSI Preprocessing: In some embodiments, to mitigate the effect of the automatic gain controller on the CSI amplitude, the measured CSI power response G(t,f) is first normalized and then Hampel-filtered to remove outliers in the time domain. In some embodiments, this CSI preprocessing is performed by the CSI preprocessing module 110 of FIG. 1.

[0133] Proximity Monitoring: The correlation of CSI power responses on adjacent subcarriers is an effective indicator of the rough distance between a moving human and a WiFi device. Figure 2 shows the correlation matrices 210, 220, and 230 between subcarriers at human-device distances of 1 m, 3 m, and 5 m, respectively. As shown in Figure 2, the closer the human movement is to the WiFi device, the higher the correlation between CSI power responses on adjacent subcarriers. Therefore, the proximity feature can be defined as: TIFF2025066624000005.tif12158 where ρ G (f n ,f n+1 ) is the correlation between adjacent subcarriers n and n+1, and f n and f n+1 is the corresponding subcarrier frequency.

[0134] Furthermore, proximity features F within a specific time window in the disclosed proximity monitoring P The slope of (F S ) can be used. When a human subject moves closer or further away from the device, F P The value of increases or decreases. As a result, the slope F S is F P This captures useful trends, allowing the system to gain valuable insight into proximity dynamics. In some embodiments, this proximity monitoring is performed by proximity monitoring module 120 of FIG.

[0135] Walking Detection: Because the proximity feature relies on dynamic reflections of WiFi signals, it is sensitive to the strength of the movement, and small movements within the same target area may result in a smaller feature value compared to larger movements. If a user enters a target area and remains stationary, the proximity feature will not consistently indicate nearby movement. Because proximity changes include walking movements, walking information can be used to determine whether a human subject has entered or exited a particular area. In some embodiments, this walking detection is performed by the walking detection module 130 of FIG. 1.

[0136] The presence or absence of human walking can be detected by identifying the gait cycle. A gait cycle is the walking pattern from the moment one foot strikes the ground to the moment the same foot strikes the ground again. Each gait can be identified by the speed during the gait, as it has a velocity peak mid-stance, followed by a velocity decrease as the foot leaves the ground, an increase in velocity during the swing phase, and continues until the foot strikes the ground again. Gait speed can be reliably estimated from a statistical model based on the ACF of CSI power as follows: TIFF2025066624000006.tif11158 where J0 is the zeroth-order Bessel function, k is the wavenumber, f is the subcarrier frequency, and g(f) is the gain coefficient of that subcarrier. Furthermore, based on the following considerations shown in Figure 3, which shows gait information 300 from the derivative of the autocorrelation function (ACF), a gait score reflecting the probability of gait presence can be defined. First, in the time domain, the human gait cycle can be estimated from the fluctuations in estimated speed. First, in the time domain, the estimated speed fluctuations can be used to infer the human gait cycle. Second, the reliability of the gait speed can be extracted based on the gait distribution of normal humans. Third, the significance of the peaks and valleys also indicates the reliability of the speed estimation. Therefore, in this system, the difference between the peaks and valleys is used as a weight to calculate the gait score.

[0137] In some embodiments, the ambulation score is formulated as follows: TIFF2025066624000007.tif12158Here, w is the difference in the prominence of velocity peaks and valleys. If neither peaks nor valleys are detected, w is equal to 0. c is the number of walking cycles per second, constrained to be between 0.5 and 1.5 for normal walking. p(v) is the probability that the observed velocity occurs if the velocity is human walking. Pedestrian walking speed follows a normal distribution with an estimated mean of 1.34 m / s and a standard deviation of 0.37 m / s, so the probability can be calculated as follows: TIFF2025066624000008.tif16158Here, Φ((v-1.34) / 0.37) is the cumulative distribution function (CDF) of a non-standard normal distribution with mean 1.34 and standard deviation 0.37.

[0138] State Transition: By combining both proximity features and gait information, it is possible to infer whether a moving human has entered the proximity area. To this end, the present teachings disclose a finite state machine (FSM) 400 containing multiple states, as shown in FIG. 4. By monitoring the trends of proximity features and gait presence, the FSM 400 improves the accuracy of the inference process. The FSM 400 includes four states: "Far State 410," "Approach State 420," "Near State 430," and "Leaving State 440."

[0139] The far state 410 is the default state of the system, where the detector or system 100 continuously monitors the ACF spectrum and calculates proximity features while outputting the probability of detecting a human walking. If the system detects an increase in proximity features along with human walking, it transitions to the close state 420; otherwise, it remains in the "far" state 410.

[0140] The FSM 400 enters the near state 420 due to the presence of human walking and an increasing proximity feature. The system continues to check the proximity feature and walking. If the proximity feature reaches a threshold indicating proximity, the FSM 400 switches to the near state 430; otherwise, it remains in the near state 420. In the absence of walking detection and a high proximity feature, the FSM 400 returns to the far state 410.

[0141] In the proximity state 430, a human subject enters a predefined proximity area, regardless of the intensity of their movement. Based on the assumption that humans only leave by walking, if the system simultaneously detects a proximity feature and a decrease in human walking, the FSM 400 transitions to the departure state 440. Otherwise, the system remains in the proximity state 430.

[0142] The departure state 440 indicates that a person is walking out of the proximity area. When the proximity feature decreases to a threshold that indicates departure, the FSM 400 switches back to the default initialization state (in this example, the far state 410). In some embodiments, all state transitions are performed by the state transition module 140 of FIG. 1.

[0143] Accurate gait extraction from CSI requires a high sounding rate, whereas proximity features do not. Therefore, downsampling is performed on the received CSI before proximity monitoring, not before gait detection, as shown in Figure 1. This downsampling reduces the computational overhead, making the system more practical for real-time applications.

[0144] Figure 5 is an example 500 showing how extracted features change over time as a person approaches the device. Initially, as the person walks towards the device, there is a noticeable fluctuation in the ACF spectrum and estimated velocity (v), accompanied by an increase in the value of the proximity feature FP. As the person enters the target area, FP P remains high and decreases if the person makes small movements (e.g., reading). Even if FP decreases, the state machine remains in the neighborhood state because no walking patterns are observed. If a walking pattern is detected and F P The departure state is triggered only when,decreases.,Integrating gait features into proximity features in the,state machine effectively enhances the robustness of the system when,humans are in quasi-static states.

[0145] In some embodiments, a pair of Tx-Rx prototypes equipped with commercially available WiFi cards can be built to implement the disclosed proximity detector. Each prototype has two omnidirectional antennas. The center frequency is set to 5.18 GHz with a bandwidth of 40 MHz. The sounding rate is set to 1500 Hz, downsampled to 30 Hz in the proximity monitoring module to reduce computational complexity.

[0146] To evaluate the performance of the disclosed system, extensive experiments can be conducted in a variety of environments, including two typical office environments and two home environments (one apartment and one house), as depicted in FIGS. 6A-6D. FIG. 6A illustrates a first office environment (Office 1 environment) 610 for wireless proximity detection. In Office 1 environment 610, a person can move across the office over time along a walking path 615. A transmitter (Tx) 611 is located in a corner of one office, while a receiver (Rx) 612 is located on a desk in another office. In some embodiments, Rx 612 can receive wireless signals transmitted by Tx 611 through a multipath channel. The wireless signals are affected by the person's movement as they move within any of the offices in Office 1 environment 610. The disclosed system can calculate a CSI time series based on the received wireless signals and extract different features, including gait information and proximity features, from the CSI time series.

[0147] 6B shows a second office environment (Office 2 environment) 620 for wireless proximity detection. In Office 2 environment 620, a person may move across the office over time along a walking path 625. Tx 621 is placed on a desk in one office, and Rx 622 is placed on a desk in another office. In some embodiments, Tx 621 and Rx 622 operate similarly to Tx 611 and Rx 612, respectively, capturing a time series of CSI or channel information for proximity detection.

[0148] 6C shows a first home environment (Home 1 environment) 630 for wireless proximity detection. In Home 1 environment 630, a person can move across rooms over time along a walking path 635. Tx 631 is located in the living room, while Rx 632 is located in the bedroom. In some embodiments, Tx 631 and Rx 632 operate similarly to Tx 611 and Rx 612, respectively, capturing a time series of CSI or channel information for proximity detection.

[0149] 6D shows a second home environment (Home 2 environment) 640 for wireless proximity detection. In Home 2 environment 640, a person can move across rooms over time along a walking path 645. Tx 641 is placed in a corner of the living room, while Rx 642 is placed on a desk in a bedroom. In some embodiments, Tx 641 and Rx 642 operate similarly to Tx 611 and Rx 612, respectively, capturing a time series of CSI or channel information for proximity detection.

[0150] During the experiment in each environment, participants walked toward the target device, stayed in close proximity to the device, performed small movements (e.g., sitting, reading a book), and then returned to the starting point. This series of actions, which could last for more than a minute, constituted an "event sample." To assess the false alarm rate, participants were allowed to move away from the device at will while maintaining it in the same fixed position. False alarms were recorded by a timestamp camera.

[0151] The following metrics can be used to evaluate the performance of the disclosed system: Figures 7A-7C show some example metrics for wireless proximity detection.

[0152] The first metric is instance-based accuracy (IA). IA measures the detection accuracy of the aforementioned "event sample," which includes a sequence of an object approaching, entering a proximity area, and leaving. In some embodiments, it is calculated as the ratio of the system's detection to the true occurrence of this sequence. In mathematical terms, IA can be calculated as follows: TIFF2025066624000009.tif18158 where N GT represents the actual number of occurrences of the event according to the truth, while N sys represents the number of detections made by the system for the same event, as shown in Figure 7A.

[0153] The second metric is duration-based accuracy (DA). In this example, the proximity area can be defined as the area within 1.5 meters of the target device. Using a timestamp camera and ground labels, we can determine the actual duration between when a person enters the target area and when they leave the area, resulting in a duration-based metric. Duration-based accuracy can be mathematically formulated as follows: TIFF2025066624000010.tif16158Here T sys is the time the system output remains in the neighborhood state. T GT is the duration of the true proximity state. Figure 7A shows the calculation of DA. In some embodiments, DA can be calculated only for samples where both the approach and departure states were successfully detected.

[0154] The third metric is responsiveness. As shown in Figure 7B, the responsiveness of a system can be assessed by measuring the time difference between the system's detection and the truth, e.g., TIFF2025066624000011.tif14158 where t GT is the timestamp of the occurrence of the actual neighborhood state, while t sys is the timestamp of the first system detection.

[0155] The fourth metric is the false alarm (FA) rate. As shown by Figure 7C, a false alarm occurs when the disclosed system incorrectly detects that a subject has entered a target area. The false alarm rate is also instance-based and can be defined as follows: TIFF2025066624000012.tif18158 where N sys FA is the number of false positive events, and the denominator N GT empty is the total number of instances when the target area is empty.

[0156] The performance of the disclosed system, tested in the four environments shown in Figures 6A-6D, is shown in Table I below. The parameters were empirically determined and are universal for all environments. The overall detection rate is 92.5%, with a false alarm rate of 1.12%. As shown in Table I, the results demonstrate a high detection rate of over 97% and a low false alarm rate of less than 3% in all environments, highlighting the effectiveness of the disclosed approach in accurately detecting proximity in a variety of scenarios. Furthermore, once a proximity instance is detected, the system exhibits a nearly perfect detection average of 98.86%, consistently aligning with the truth. The overall detection latency of 0.825 seconds also demonstrates responsive system detection. Table 1 TIFF2025066624000013.tif46153

[0157] In some embodiments, the walking paths tested were longer than 5 meters to ensure stable walking periods. To evaluate how the length of the walking path affects performance, we can collect more data on different walking path lengths, including 3 meters, 6 meters, and 9 meters, by adjusting the starting point in the Office 1 environment 610. The accuracy results indicate that shorter path lengths result in lower IA because walking patterns may not be sufficiently stable during transition periods. Furthermore, because the DA value is calculated when the approach-departure state is successfully recognized, the DA value approaches 100% when the approach-departure state is identified.

[0158] FIG. 8 shows a flowchart of an example method 800 for wireless proximity detection according to some embodiments of the present disclosure. In various embodiments, method 800 may be performed by the systems disclosed above. At operation 802, a wireless signal is transmitted by a first wireless device through a wireless multipath channel in a venue. At operation 804, the wireless signal is received by a second wireless device through the wireless multipath channel. The received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel and the movement of objects in the venue. At operation 806, time-series channel information (TSCI) of the wireless multipath channel is obtained based on the received wireless signal. At operation 808, first operational statistics are calculated based on a time-domain correlation of the channel information (CI) in the TSCI. At operation 810, second operational statistics are calculated based on a frequency-domain correlation of the CI in the TSCI. At operation 812, proximity of an object to a target in the venue is detected based on the first operational statistics and the second operational statistics.

[0159] FIG. 9 is a flowchart illustrating detailed operations 900 for wireless proximity detection according to some embodiments of the present disclosure. In some embodiments, operation 900 may be performed as part of operation 808 of FIG. 8. In operation 910, a feature of each CI in the TSCI is calculated. The feature includes one of the following: magnitude, phase, magnitude of a component of the CI, phase of a component of the CI, magnitude squared, or a function of magnitude. In operation 920, a time-domain autocorrelation function (ACF) is calculated based on the features of each CI in a first sliding time window of the TSCI. The time-domain ACF is composed of multiple time-domain correlations. In operation 930, first motion statistics are calculated based on the time-domain ACF. In some embodiments, the second motion statistics in operation 810 of FIG. 8 may similarly be calculated based on the frequency-domain ACF.

[0160] The order of the actions in any one of the figures may be varied according to various embodiments of the present teachings.

[0161] In some embodiments of the present teachings, the proximity of an object (e.g., a human) to a target within a venue can be calculated / determined / detected based on wireless channel information (e.g., CI, CSI, CIR, CFR) obtained from wireless signals (e.g., IEEE 802.11 signals, 802.11bf signals, mobile communications / 4G / 5G / 6G / 7G / 8G / 9G / 10G signals, UWB, wireless sounding signals, standard-compliant signals) communicated between a Type 1 (TX) wireless device and a Type 2 (RX) wireless device.

[0162] At the venue, targets (e.g., floors, rooms, neighborhoods, areas, locations, devices, appliances, second objects) may be determined. Targets may be in the vicinity of either Type 1 or Type 2 devices. Targets can be either Type 1 or Type 2 devices. Targets can be smart / WiFi-enabled appliances / furniture / fixtures such as TVs, speakers, plugs, lighting / lights, clocks, security cameras, doorbells, vacuum cleaners, etc.

[0163] Proximity (e.g., including proximity state / amount / tag / state / class / index / score) may be calculated based on both time-domain correlation and frequency-domain correlation of CIs in a time series of CIs (TSCI). A first motion statistic, e.g., the walking score F_g in Equation (6), may be calculated based on the time-domain correlation of CIs in a sliding time window. A second motion statistic, e.g., F_p in Equation (4), may be calculated based on the frequency-domain correlation of CIs in the TSCI. F_s may be the slope (first derivative) of the second-order motion statistic within a time window. A positive F_s may suggest that the second motion statistic (or its magnitude) is increasing over time, while a negative F_s may mean that it is decreasing over time. A large positive F_s suggests that the second-order motion statistic is increasing at a fast pace, while a small positive F_s suggests that it is increasing slowly.

[0164] The first operating statistic (MS) may be calculated for multiple time instances forming a first time trend. The first MS may be calculated at regular time intervals at a first (sampling) frequency / rate, which may be 0.0001 / 0.001 / 0.01 / 0.1 / 10 / 100 / 1000 / 10000 / 100000 Hz or other values. The initial (sampling) frequency / rate may be changed over time.

[0165] The second operating statistics may be calculated for multiple time instances forming a second time trend. The second MS may be calculated at regular time intervals at a second (sampling) frequency / rate, which may be 0.0001 / 0.001 / 0.01 / 0.1 / 10 / 100 / 1000 / 10000 / 100000 Hz, or other values. The second sampling frequency / rate may be the same as or different from the first sampling rate. The second sampling rate may be changed over time.

[0166] The proximity (e.g., one of four states of a finite state machine) may be calculated for multiple time instances forming a third time trend. The proximity may be calculated at regular time intervals at a third (sampling) frequency / rate of 0.0001 / 0.001 / 0.01 / 0.1 / 10 / 100 / 1000 / 10000 / 100000 Hz, or other values. The third sampling rate may be the same as or different from the first or second sampling rate. The third sampling rate may change over time.

[0167] The first MS, the second MS, and the proximity may be calculated repeatedly at their respective timings, for example, periodically at their respective rates.

[0168] The TSCI may be sampled at a high sampling frequency (e.g., 10,000 Hz). The relevant (time-domain or frequency-domain) correlation of one of the MSs (e.g., a first MS) and the CI may be calculated based on the entire TSCI at the high sampling frequency. The TSCI may be resampled / subsampled to a lower sampling frequency (e.g., 100 Hz). The relevant (frequency-domain or time-domain) correlation of another one of the MSs (e.g., a second MS) and the CI may be calculated based on the resampled / subsampled CI of the TSCI.

[0169] The proximity may be calculated at a rate equal to or less than the low sampling frequency (e.g., 1 Hz). Both the first MS and the second MS may be calculated (e.g., at 1 Hz) at (or shortly before) the time when the proximity needs to be calculated. In this case, the first MS is calculated based on a first time window (e.g., 0.1 seconds) of the CI of the TSCI, and the second MS is calculated based on a second time window (e.g., 1 second) of the CI resampled / subsampled from the TSCI.

[0170] The first motion statistic (MS) may be an ambulation score as defined in Equation (6). The first MS (e.g., ambulation score) may be calculated for a sliding time window based on a time-domain autocorrelation function calculated using the CIs within the sliding time window of the TSCI. A feature (CI-feature) for each CI may be obtained. The CI feature may be a magnitude, a magnitude squared, a function of magnitude, or a phase. For two adjacent CIs, the correlation between them may simply be the correlation of two corresponding CI features. If each CI has multiple components (e.g., CSI, CIR, CFR), a component-feature (e.g., a magnitude, a magnitude squared, a function of magnitude, or a phase) may be obtained for each component. The corresponding CI feature may be a vector of component-features. For two adjacent CIs, the correlation between them may be the dot product of two corresponding vectors of component features. A time-domain ACF may be calculated based on the CI features of the CIs within the sliding time window.

[0171] At least one feature point or ACF feature (e.g., local / global maximum / peak, local / global minimum, zero crossing, mean crossing, inflection point) may be calculated / identified / discovered for either the time-domain ACF or a function of the time-domain ACF (e.g., first derivative / slope of the time-domain ACF). A first MS may be calculated based on the at least one feature point (e.g., first local maximum, first local minimum).

[0172] There may be two feature points (e.g., first local maximum + first local minimum, first local maximum + second local minimum, first local minimum + second local maximum, second local minimum + second local maximum, first local maximum + first zero crossing, first local minimum + first zero crossing, first local maximum + first inflection point, first local minimum + first inflection point, first zero crossing + first inflection point, etc.), and the first MS may be calculated based on a function of the two feature points (sum, difference, weighted sum, weighted difference, product, quotient, weighted product, weighted quotient, function of functions, sum of logarithms, etc.).

[0173] There may be multiple feature points (e.g., first local maximum + first local minimum + second local maximum + second local minimum, first local maximum + first zero crossing + first local minimum, first local maximum + first inflection point + first local minimum, etc.) The first MS may be calculated based on a function of the multiple feature points (e.g., sum, difference, weighted sum, weighted difference, product, quotient, weighted product, weighted quotient, function of functions, sum of logarithms).

[0174] In a special case, two feature points may consist of a local maximum and a local minimum (e.g., the first local maximum, the first local minimum). The first MS is calculated based on the difference between the local maximum and the local minimum. The "difference" is "w" in Equation (6).

[0175] If a local maximum or minimum cannot be found exactly, the difference may be defined / calculated as zero.

[0176] STI (e.g. speed) is calculated based on two feature points.

[0177] A probability score, e.g., “p(v)” in equation (6), is calculated for the STI based on the pdf, which is provided / calculated / trained in a training phase / session.

[0178] PDFs can be used to compute probability scores. A PDF may be a mixture of PDFs, at least one of which is a generalized Gaussian (e.g., Laplacian for order=1, Gaussian for order=2). A generalized Gaussian may be Gaussian. A mixture of PDFs may be Gaussian. A mixture PDF is a PDF of the form w_1*pdf_1+w_2*pdf_2+...+w_n*pdf_n, where w_i>=0, w_1+w_2+...+w_n=1. Each pdf_i is a PDF, i.e., a non-negative PDF whose area (integral) equals 1. In other words, a mixture of PDFs is a weighted sum of PDFs with non-negative weights that add up to 1.

[0179] The first motion statistic may be monotonically non-decreasing (eg, increasing) with respect to the probability score, or may be monotonically non-decreasing (eg, increasing) with respect to the difference between two feature points.

[0180] The sliding time window for calculating the frequency domain correlation is different from the sliding time window for calculating the time domain correlation.

[0181] Each CI (eg, CI, CSI, CIR, CFR) may have a representation (eg, a frequency representation, or a transformation to a frequency representation) that includes multiple frequency components, each with a corresponding frequency index.

[0182] Multiple (component-dependent or frequency-dependent) k-component correlations may be calculated for each component. The k-component correlation of an n-th component may be the correlation between the n-th component and the (n+k)-th component (or the n-th component). For k=1, the correlation may be calculated using the immediate neighbors in equation (4), e.g., f_n, f_{n+1} (e.g., rho_G(f_n, f_{n+1})). The frequency-domain correlation may be calculated as an aggregation (e.g., average) of the k-component correlations across all n components.

[0183] The feature may be calculated for each component of the CI. The k-component correlation of the n-th component may be / include the correlation between the feature of the n-th component and the feature of the n+k-th component (or the nk-th component).

[0184] Multiple k-component correlations may be calculated, each multiple associated with a respective value of k. For example, the associated values ​​of k may include k1, k2, k3, .... For each value of k (e.g., k1, k2, k3, ...), at least one aggregation may be calculated. For example, when k=k1, one aggregation may be the arithmetic mean and another aggregation may be the geometric mean. When k=k2, one aggregation may be the weighted mean, another aggregation may be the trimmed mean, and yet another aggregation may be the harmonic mean. The frequency-domain correlation may be based on an aggregation of all aggregations for all values ​​of k.

[0185] The proximity detection of an object to (or relative to) a target may operate according to a finite state machine (FSM). The FSM may have four states: “far,” “close,” “near,” and / or “distant.” The initial state is determined to be one of the four states. At any time t, the state at the next time may be determined based on the current state at the current time, the current value of the first motion statistic in Equation (6) based on the time-domain correlation / time-domain ACF, e.g., the walking score F_g, and the current value of the second motion statistic in Equation (4) based on the frequency-domain correlation, e.g., F_p. F_s may be the slope (or derivative) of the second motion statistic (e.g., F_p) in the time window.

[0186] In the initialization procedure, the initial state may be determined to be one of four states (e.g., far). The current state may be initialized as far. In the FSM, the state transition behavior is as described in sections 21 through 25 below.

[0187] In some examples, the target is near either a Type 1 device or a Type 2 device. In some examples, the target is either a Type 1 device or a Type 2 device.

[0188] The following numbered sections provide examples of wireless proximity detection.

[0189] Item 1. A wireless sensing method / device / system / software, comprising: transmitting a wireless signal by a Type 1 heterogeneous wireless device through a wireless multipath channel in a venue; receiving the wireless signal by a Type 2 heterogeneous wireless device, where the received wireless signal is different from the wireless signal transmitted due to the wireless multipath channel and the movement of an object in the venue; obtaining time-series channel information (TSCI) of the wireless multipath channel based on the received wireless signal; calculating first operational statistics based on time-domain correlation of CIs of the TSCI; calculating second operational statistics based on frequency-domain correlation of CIs of the TSCI; and detecting proximity of the object to a target in the venue based on the first operational statistics and the second operational statistics.

[0190] Clause 2. The wireless sensing method / apparatus / system / software of clause 1, including calculating a third time trend of object proximity to the target based on the first time trend of the first operational statistic and the second time trend of the second operational statistic.

[0191] Clause 3. The wireless sensing method / apparatus / system / software of clause 1 or 2, including repeatedly calculating, at a first repetition rate, each first operational statistic based on a first sliding window for each of the TSCIs, repeatedly calculating, at a second repetition rate, each second operational statistic based on a second sliding window for each of the TSCIs, and repeatedly calculating proximity at a third repetition rate.

[0192] Clause 4. The wireless sensing method / apparatus / system / software of clauses 1-3, including: calculating a time-domain autocorrelation function (ACF) based on the CI in a first sliding time window of the TSCI, the time-domain ACF including a plurality of time-domain correlations; and calculating a first operating statistic based on the ACF.

[0193] Clause 5. The wireless sensing method / apparatus / system / software of clause 4, including: computing a feature for each CI of the TSCI, the feature including one of magnitude, phase, magnitude of a component of the CI, phase of a component of the CI, magnitude squared, or a function of magnitude; and computing a time-domain ACF based on the feature for each CI within a sliding time window of the TSCI.

[0194] Clause 6. The wireless sensing method / apparatus / system / software of clause 4 or 5, comprising: calculating at least one feature point of the time-domain ACF or a function of the time-domain ACF; and calculating a first operating statistic based on the at least one feature point, wherein the at least one feature point comprises at least one of the following: a global maximum, a global minimum, a bounded maximum, a bounded minimum, a maximum bounded to a positive argument of the ACF or a function of the ACF, a maximum bounded to a negative argument of the ACF or a function of the ACF, a minimum bounded to a positive argument of the ACF or a function of the ACF, a minimum bounded to a negative argument of the ACF or a function of the ACF, a local maximum, a local minimum, a first local maximum, a first local minimum, a second local maximum, a second local minimum, a third local maximum, a third local minimum, a variable an inflection point, a zero-crossing point, a first inflection point, a first zero-crossing point, a second inflection point, a second zero-crossing point, a third inflection point, or a third zero-crossing point; wherein the function of the time-domain ACF includes at least one of the following: a linear function, a piecewise linear function, a nonlinear function, a polynomial function, an exponential function, a logarithmic function, a trigonometric function, a transcendental function, a differential function, a first derivative, a second derivative, a third derivative, an integral function, a single integral, a double integral, a triple integral, an absolute function, a magnitude function, an index function, a threshold function, a quantization function, a function obtained by filtering the time-domain ACF, wherein the filtering includes at least one of low-pass filtering, band-pass filtering, high-pass filtering, smoothing filtering, or weighted averaging.

[0195] Clause 7. The wireless sensing method / apparatus / system / software of clauses 4 through 6, including calculating a plurality of feature points, either a time-domain ACF or a function of the time-domain ACF, and calculating a first operating statistic based on another function of the plurality of features.

[0196] Clause 8. The wireless sensing method / apparatus / system / software of clauses 4 through 7, including calculating a local maximum as a first feature point of either the time-domain ACF or a function of the time-domain ACF, calculating a local minimum as a second feature point of either the time-domain ACF or a function of the time-domain ACF, calculating a difference by subtracting the local minimum from the local maximum, and calculating a first operating statistic based on the difference.

[0197] Clause 9. The wireless sensing method / apparatus / system / software of clause 8, including the local maximum being the first local maximum with a positive argument and the local minimum being the first local minimum with a positive argument.

[0198] Clause 10. The wireless sensing method / apparatus / system / software according to clause 8 or 9, further comprising: if at least one of the first feature point or the second feature point cannot be determined, calculating a difference between the two feature points as zero.

[0199] Clause 11. The wireless sensing method / apparatus / system / software of clauses 8 through 10, including: calculating spatial-temporal information (STI) based on a third characteristic point of either a time-domain ACF or a function of the time-domain ACF, the STI including one of position, distance, velocity, and acceleration; calculating a probability score based on the STI; and calculating a first motion statistic based on both the probability score and a difference between the first and second characteristic points.

[0200] Clause 12. The wireless sensing method / apparatus / system / software of clause 11, comprising calculating a probability score based on a probability density function (pdf).

[0201] Clause 13. The wireless sensing method / apparatus / system / software of clause 12, wherein the probability density function comprises a mixture of pdfs, and at least one pdf is a generalized Gaussian.

[0202] Clause 14. The wireless sensing method / apparatus / system / software of clauses 11 to 13, wherein the first operational statistic is monotonically non-decreasing with respect to the probability score and with respect to the difference between the two feature points.

[0203] Clause 15. The wireless sensing method / apparatus / system / software of clauses 1 through 14, including calculating a frequency domain correlation based on the CI in a second sliding time window of the TSCI.

[0204] Clause 16. The wireless sensing method / apparatus / system / software of clauses 1 through 15, wherein each CI includes a representation having multiple components, each component associated with a frequency index; calculating multiple k-component correlations, each k-component correlation being a correlation between a respective one of the components of the CI and a respective one of adjacent components of the CI, wherein the frequency indexes of two respective components of the CI differ by k, where k is an integer greater than 0; and calculating a frequency-domain correlation based on an aggregation of the multiple k-component correlations, wherein any aggregation includes at least one of the following: sum, weighted sum, average, weighted average, trimmed mean, product, weighted product, arithmetic mean, geometric mean, harmonic mean, or another aggregation of any of the above.

[0205] Clause 17. The wireless sensing method / apparatus / system / software of clause 16, comprising: calculating a feature for each component of the CI, the feature comprising one of magnitude, phase, magnitude squared, or a function of magnitude; and calculating each k-component correlation based on the feature for each component and the features for each adjacent component.

[0206] Clause 18. The wireless sensing method / apparatus / system / software of clause 16 or 17, comprising: computing a first plurality of k-component correlations for a first value of k; computing at least one first aggregation of a first of the first plurality of k-component correlations for the first value of k; computing a second plurality of k-component correlations for a second value of k; computing at least one second aggregation of the second plurality of k-component correlations for the second value of k; and computing a frequency domain correlation based on a third aggregation of the at least one first aggregation and the at least one second aggregation.

[0207] Clause 19. The wireless sensing method / apparatus / system / software of clauses 1 through 18, including: determining at least four states of proximity of an object to a target, including the following states: a far state indicating the object is far away from the target, a close state indicating the object is moving toward the target, a near state indicating the object is near the target, and a distant state indicating the object is moving away from the target; determining an initial state, which is one of the at least four states; and determining a next state at a next time based on a current state at a current time, a first operating statistic, and a second operating statistic.

[0208] Clause 20. The wireless sensing method / apparatus / system / software of clause 19, including determining the initial state to be a distant state.

[0209] Clause 21. The wireless sensing method / apparatus / system / software according to clause 19 or 20, including: if the current state at the current time is far: if the first operating statistic (F_g) is greater than a first threshold and the slope of the second operating statistic (F_s) is greater than a second threshold, transition from the far state to the close state; otherwise, remain in the far state.

[0210] Clause 22. The wireless sensing method / apparatus / system / software of clauses 19 to 21, including: if the current state is an approach state at the current time, if the second operation statistic (F_p) is greater than a third threshold, transition from the approach state to the proximity state; if the first operation statistic is greater than a fourth threshold and the slope of the second operation statistic is less than a fifth threshold (wherein the fifth threshold is a negative value), transition from the approach state to the distance state; if the first operation statistic is greater than a sixth threshold and the slope of the second operation statistic is greater than a seventh threshold, remain in the approach state; otherwise, transition from the approach state to the distance state.

[0211] Clause 23. The wireless sensing method / apparatus / system / software of clauses 19 to 22, including: if the current state at the current time is proximity: if the first operating statistic is greater than an eleventh threshold and the slope of the second operating statistic is less than a twelfth threshold, where the twelfth threshold is negative, transition from the proximity state to the departure state; otherwise, remain in the proximity state.

[0212] Clause 24. The wireless sensing method / apparatus / system / software of clauses 19 to 23, including: if the current state at the current time is a Distant state: if the second operating statistic is less than a thirteenth threshold, transition from the Distant state to a Far state; if the first operating statistic is greater than a fourteenth threshold and the slope of the second operating statistic is greater than a fifteenth threshold, transition from the Distant state to an Approaching state; and otherwise remain in the Distant state.

[0213] Clause 25. A wireless sensing method / device / system / software of clauses 1 through 24, including a target being in proximity to either a Type 1 device or a Type 2 device.

[0214] Clause 26. The wireless sensing method / device / system / software of clauses 1 through 25, wherein the target is either a Type 1 device or a Type 2 device, and the proximity of the target is relative to either the Type 1 device or the Type 2 device.

[0215] The following numbered sections provide examples of wireless proximity detection.

[0216] Section 3812_1. A method for wireless sensing, comprising: transmitting, by a first wireless device, a wireless signal through a wireless multipath channel in a venue; receiving, by a second wireless device, the wireless signal through the wireless multipath channel, where the received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel and motion of an object in the venue; obtaining time-series channel information (TSCI) of the wireless multipath channel based on the received wireless signal; calculating first operational statistics based on a time-domain correlation of the channel information (CI) in the TSCI; calculating second operational statistics based on a frequency-domain correlation of the CI in the TSCI; and detecting proximity of an object to a target in the venue based on the first operational statistics and the second operational statistics.

[0217] Clause 3812_2. The method of Clause 3812_1, further including calculating a third time trend of the object's proximity to the target based on a first time trend of a first operational statistic and a second time trend of the second operational statistic.

[0218] Clause 3812_3. The method of clause 3812_1, further including: repeatedly calculating first motion statistics at a first repetition rate, each first motion statistic being calculated based on a respective first sliding window of the TSCI; repeatedly calculating second motion statistics at a second repetition rate, each second motion statistic being calculated based on a respective second sliding window of the TSCI; and repeatedly calculating a proximity of the object to the target at a third repetition rate.

[0219] Clause 3812_4. The method of clause 3812_1, further including: calculating a time-domain autocorrelation function (ACF) based on the CI in a first sliding time window of the TSCI, the time-domain ACF including a plurality of time-domain correlations; and calculating a first motion statistic based on the time-domain ACF.

[0220] Clause 3812_5. The method of clause 3812_4, further including: calculating a feature of each CI in the TSCI, the feature including one of a magnitude, a phase, a magnitude of a component of the CI, a phase of a component of the CI, a magnitude squared, or a function of the magnitude; and calculating a time-domain ACF based on the feature of each CI within a first sliding time window of the TSCI.

[0221] Clause 3812_6. The method of clause 3812_4, further comprising: calculating at least one feature point of the time-domain ACF or a function of the time-domain ACF; and calculating a first operating statistic based on the at least one feature point, wherein the at least one feature point includes at least one of the following: a global maximum, a band minimum, a restricted maximum, a restricted minimum, a maximum restricted to a positive argument of the ACF or a function of the ACF, a maximum restricted to a negative argument of the ACF or a function of the ACF, a minimum restricted to a positive argument of the ACF or a function of the ACF, a local maximum, a local minimum, a first local maximum, a first local minimum, a second local maximum, a second local minimum, a third local maximum, a third local minimum, an inflection point, a zero-crossing point, a mean-crossing point, a first inflection point, a first zero. crossing point, first average crossing point, second inflection point, second zero crossing point, second average crossing point, third inflection point, third zero crossing point, or third average crossing point; where the function of the time-domain ACF includes at least one of the following: a linear function, a piecewise linear function, a nonlinear function, a polynomial function, an exponential function, a logarithmic function, a trigonometric function, a transcendental function, a differential function, a first differential, a second differential, a third differential, an integral function (single integral, double integral, triple integral), an absolute function, a magnitude function, an index function, a threshold function, a quantization function, or a function obtained by filtering the time-domain ACF, where the filtering includes at least one of the following: low-pass filtering, band-pass filtering, high-pass filtering, smoothing filtering, or weighted averaging.

[0222] Clause 3812_7. The method of clause 3812_6, further including calculating a plurality of feature points of either the time-domain ACF or a function of the time-domain ACF, and calculating a first motion statistic based on an additive function of the plurality of feature points.

[0223] Clause 3812_8. The method of clause 3812_7, further including calculating a local maximum as a first feature point of either the time-domain ACF or a function of the time-domain ACF, calculating a local minimum as a second feature point of either the time-domain ACF or a function of the time-domain ACF, calculating a difference between the first and second feature points by subtracting the local minimum from the local maximum, and calculating a first motion statistic based on the difference.

[0224] Item 3812_9. The method of Item 3812_8, wherein the local maximum is the first local maximum for which the argument is positive, and the local minimum is the first local minimum for which the argument is positive.

[0225] Section 3812_10. The method of Section 3812_8, wherein if at least one of the first feature point or the second feature point cannot be determined, the difference between the first feature point and the second feature point is calculated as zero.

[0226] Clause 3812_11. The method of clause 3812_8, further including: calculating spatial-temporal information (STI) based on a third feature point of either the time-domain ACF or a function of the time-domain ACF, where the STI includes one of position, distance, velocity, or acceleration; calculating a probability score based on the STI; and calculating a first motion statistic based on both the probability score and a difference between the first and second feature points.

[0227] Item 3812_12. The method of Item 3812_11, further including calculating a probability score based on a probability density function (pdf).

[0228] Item 3812_13. The method of Item 3812_12, wherein the probability density function comprises a mixture of pdfs, and at least one of the pdfs is a generalized Gaussian pdf.

[0229] Item 3812_14. The method of Item 3812_11, wherein the first operating statistic is monotonically non-decreasing with respect to the probability score and with respect to the difference between the first and second feature points.

[0230] Clause 3812_15. The method of clause 3812_1, further comprising calculating a frequency domain correlation based on the CI in a second sliding time window of the TSCI.

[0231] Clause 3812_16. The method of Clause 3812_15, including computing a plurality of k-component correlations, each CI including a representation having a plurality of components, each of the plurality of components associated with a respective frequency index, each of the plurality of k-component correlations being a correlation between each of the components of the CI and each of the adjacent components of the CI, wherein the frequency indexes of each of the components of the CI and each of the adjacent components of the CI differ by k, where k is an integer greater than 0; and computing a frequency-domain correlation based on an aggregation of the plurality of k-component correlations, the aggregation including at least one of a sum, a weighted sum, an average, a weighted average, a trimmed mean, a product, a weighted product, an arithmetic mean, a geometric mean, a harmonic mean, or another aggregation of any of the foregoing.

[0232] Clause 3812_17. The method of clause 3812_16, further including: for each k-component correlation for each CI: calculating a feature of each component of each CI, the feature including one of magnitude, phase, magnitude squared, or a function of magnitude, and calculating each k-component correlation based on the feature of each component and the feature of each adjacent component.

[0233] Clause 3812_18. The method of clause 3812_16, further including: computing a first plurality of k1-component correlations, where k1 is a first integer greater than zero, computing at least one first tally of the first plurality of k1-component correlations; computing a second plurality of k2-component correlations, where k2 is a second integer greater than zero, computing at least one second tally of the second plurality of k2-component correlations; and computing a frequency-domain correlation based on the at least one first tally and a third tally of the at least one second tally.

[0234] Clause 3812_19. The method of clause 3812_1, including determining at least four states of proximity of the object to the target, the at least four states including a far state indicating the object is far away from the target, a close state indicating the object is moving toward the target, a near R state indicating the object is near the target, and a away state indicating the object is moving away from the target; determining one of the at least four states as an initial state; and determining a next state at a next time based on a current state at a current time, first operating statistics, and second operating statistics.

[0235] Section 3812_20. The method of Section 3812_19, further comprising determining the initial state to be a distant state.

[0236] Section 3812_21. The method of Section 3812_19, if the current state at the current time is the far state: transitioning from the far state to the close state in response to determining that a condition is met, the condition being (a) the first operating statistic being greater than a first threshold and (b) the slope of the second operating statistic being greater than a second threshold; and remaining in the far state in response to determining that the condition is not met.

[0237] Section 3812_22. The method of Section 3812_19, further including, if the current state at the current time is the Approaching State: transitioning from the Approaching State to the Proximity State in response to determining that a first condition is met, the first condition being that the second operating statistic is greater than a third threshold; transitioning from the Approaching State to the Disengaging State in response to determining that the second condition is met, the second condition including (a) the first operating statistic being greater than a fourth threshold and (b) a slope of the second operating statistic being less than a fifth threshold, where the fifth threshold is a negative number; remaining in the Approaching State in response to determining that a third condition is met, the third condition including (c) the first operating statistic being greater than a sixth threshold and (d) a slope of the second operating statistic being greater than a seventh threshold; and transitioning from the Approaching State to the Disengaging State in response to determining that none of the first, second, and third conditions are met.

[0238] Clause 3812_23. The method of Clause 3812_19, further including, if the current state at the current time is a proximity state: transitioning from the proximity state to a leaving state in response to determining that a condition is met, the condition being (a) the first operating statistic being greater than an eighth threshold and (b) the slope of the second operating statistic being less than a ninth threshold, the ninth threshold being a negative number; and remaining in the proximity state in response to determining that the condition is not met.

[0239] Section 3812_24. The method of Section 3812_19, further including, if the current state at the current time is the Departure state: transitioning from the Departure state to the Far state in response to determining that a first condition is met, the first condition including the second operating statistic being less than a tenth threshold; transitioning from the Departure state to the Approach state in response to determining that a second condition is met, the second condition including (a) the first operating statistic being greater than an eleventh threshold and (b) a slope of the second operating statistic being greater than a twelfth threshold; and remaining in the Departure state in response to determining that neither the first nor the second condition is met.

[0240] Clause 3812_25. The method of clause 3812_1, wherein the target is proximate to one of the first wireless device or the second wireless device.

[0241] Clause 3812_26. The method of clause 3812_1, wherein the target is either a first wireless device or a second wireless device, and the proximity of the target is relative to either the first wireless device or the second wireless device.

[0242] Item 3812_27. A system for wireless sensing, comprising: a first wireless device configured to transmit a wireless signal through a wireless multipath channel in a venue; a second wireless device configured to receive the wireless signal through the wireless multipath channel, where the received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel and motion of an object in the venue; and a processor configured to: obtain time-series channel information (TSCI) of the wireless multipath channel based on the received wireless signal; calculate first operating statistics based on a time-domain correlation of the channel information (CI) in the TSCI; calculate second operating statistics based on a frequency-domain correlation of the CI in the TSCI; and detect proximity of an object to a target in the venue based on the first operating statistics and the second operating statistics.

[0243] Clause 3812_28. The system of clause 3812_27, wherein the processor is further configured to repeatedly calculate first motion statistics at a first repetition rate, each first motion statistic being calculated based on a first sliding window for a respective one of the TSCIs, repeatedly calculate second motion statistics at a second repetition rate, each second motion statistic being calculated based on a second sliding window for a respective one of the TSCIs, and repeatedly calculate the proximity of the object to the target at a third repetition rate.

[0244] Clause 3812_29. The system of clause 3812_27, further including: calculating a feature of each CI in the TSCI, the feature including one of magnitude, phase, magnitude of a component of the CI, phase of a component of the CI, magnitude squared, or a function of magnitude; calculating a time-domain autocorrelation function (ACF) based on the feature of each CI in a first sliding time window of the TSCI, the time-domain ACF including a plurality of time-domain correlations; and calculating a first operating statistic based on the time-domain ACF.

[0245] Item 3812_30. An apparatus for wireless sensing, comprising: a receiver configured to receive a wireless signal transmitted by a transmitter over a wireless multipath channel in a venue, wherein the received wireless signal differs from a transmitted wireless signal due to the wireless multipath channel and movement of an object in the venue; a memory having an instruction set stored therein; and a processor communicatively coupled to the memory and configured to: obtain time-series channel information (TSCI) of the wireless multipath channel based on the received wireless signal; calculate first operating statistics based on a time-domain correlation of the channel information (CI) in the TSCI; calculate second operating statistics based on a frequency-domain correlation of the CI in the TSCI; and detect proximity of an object to a target in the venue based on the first operating statistics and the second operating statistics.

[0246] In recent years, indoor intelligent perception systems have attracted much attention. However, accurately detecting human presence is challenging in the presence of non-human objects such as pets, robots, and electrical appliances, limiting the practicality of these systems for widespread use, as depicted in Figure 10.

[0247] The present teachings disclose a novel system, called "WI-MOID," a WiFi-based human-nonhuman motion identification system capable of accurately identifying various human and nonhuman subjects through walls. In some embodiments, the system uses a set of general-purpose WiFi transceivers to passively and unobtrusively identify moving humans and various nonhuman objects without requiring any devices or restricting their motion. WI-MOID leverages a novel multipath model based on statistical electromagnetic wave theory to detect moving objects and extract physically and statistically explainable features of their motion, accurately distinguishing between human and various nonhuman motions through walls, even in complex environments. Furthermore, WI-MOID is suitable for edge devices, requires minimal computing resources and storage, and is environment-independent, allowing it to be easily deployed in new environments with minimal effort. The performance of WI-MOID is validated in four different buildings with various moving objects, including pets, a vacuum cleaner robot, humans, and electric fans. The results show that it achieves high accuracy and low false alarm rate in identifying human and non-human motion, including high accuracy in unknown environments without model tuning, demonstrating its robustness for ubiquitous use.

[0248] In some embodiments, the disclosed system utilizes ubiquitous WiFi signals and operates unobtrusively and contactlessly, eliminating the need for additional instrumentation or environmental constraints. In contrast to existing systems that require pets, robots, and humans to move along predefined paths, directions, or regions, WI-MOID can detect freely moving subjects, allowing them to walk, turn, stand, or stand still without predefined restrictions. WI-MOID automatically detects subject motion and extracts context-independent features, allowing it to function effectively in diverse environments without requiring additional training or parameter tuning.

[0249] Realizing such a flexible system is not easy and several major challenges must be addressed. The first challenge is deriving motion signatures in both line-of-sight (LOS) and non-line-of-sight (NLOS) settings using only a single-link WiFi signal. Velocity patterns, a key component of motion signatures, present unique challenges. While various WiFi-based velocity estimation approaches, such as those using Doppler frequency shift (DFS), have been attempted, they typically require multiple links or are only effective under line-of-sight (LOS) conditions. In some embodiments, the system employs a rich scattering model to overcome this challenge by treating targets as multiple scatterers. Second, developing a system that operates effectively in unknown contexts without requiring additional training or parameter tuning is a substantial challenge. Most existing WiFi-based activity recognition methods rely on features extracted directly from raw CSI and are susceptible to environmental changes, resulting in poor performance in new environments without retraining. To address this issue, our system leverages the autocorrelation function (ACF), a second-order statistic of CSI, to extract only features caused by changes in the multipath profile. Unlike raw CSI, ACF is not affected by stationary objects and only reflects the behavior of moving objects, making it suitable for behavior recognition in a variety of environments. Third, accurately recognizing pet behavior in real-world environments is challenging, especially when pets' movement paths, ranges, and actions are not restricted. Previous research aimed at pet behavior recognition often makes unrealistic assumptions about pet behavior, such as assuming a specific path or limited area, considering only walking activities, or requiring pets to stay quietly near the device for long periods of time. In this paper, we present a novel approach that combines physical and statistical features to classify human and non-human behaviors, and further enhances classification using a hidden Markov model (HMM)-based state machine that leverages temporal information to improve accuracy. By incorporating temporal information, WI-MOID can accurately and reliably distinguish between pet and human behaviors, even when pets are moving freely and performing different activities.

[0250] Large-scale deep learning models, while powerful, are often too resource-intensive for these devices. WI-MOID addresses this issue with a lightweight Support Vector Machine (SVM) model that leverages effective features with physical and statistical significance, requiring less computational power and fewer parameters, making it suitable for edge devices. Its minimal latency has been proven in real-world tests, and it has been confirmed to operate in real time on resource-limited devices.

[0251] In some embodiments, WI-MOID can be implemented using commodity-off-the-shelf (COTS) WiFi devices. Extensive experiments were conducted in four typical indoor environments: an apartment, a townhouse, a single-family home, and a typical office. Most data was acquired under non-linear operating conditions to ensure real-world relevance. WI-MOID demonstrates impressive performance, with high validation accuracy and high test accuracy in unknown environments. This high accuracy, combined with its environment-independent features, positions WI-MOID as a promising solution for detecting, distinguishing, and monitoring human and non-human activity across diverse and extensive contexts. WI-MOID is expected to find applications in areas such as security, home automation, and activity recognition.

[0252] Thus, the present teachings disclose a motion detection and recognition system capable of accurately monitoring and recognizing human and non-human objects. By extracting a wide range of physically plausible and statistically relevant features, WI-MOID can detect and recognize humans, pets, robots, and other electrical appliances, regardless of their position, orientation, or environmental context. Furthermore, the present teachings disclose an HMM-based state machine that incorporates temporal information to correct misclassification, enhancing the system's ability to accurately recognize complex motions and improving overall accuracy and stability in practical applications. Furthermore, the present teachings disclose a motion classification scheme for distinguishing between human and non-human objects without relying on data-intensive deep learning techniques, and compiled a comprehensive CSI dataset of human and various non-human motions. We built a prototype of WI-MOID using readily available commercial off-the-shelf (COTS) WiFi devices and conducted experiments in various indoor environments. Results demonstrated high accuracy and suitability for deployment in a variety of environments. Furthermore, WI-MOID can be implemented to operate in real time on a variety of edge devices with limited computing resources and storage, demonstrating its readiness for ubiquitous deployment.

[0253] In some embodiments, the disclosed system not only performs motion classification of human and non-human objects, but also discloses a novel state machine based on hidden Markov chains to classify non-human objects and consider historical information and temporal characteristics of the output, enabling more accurate and stable recognition of moving objects in the environment. Based on performance evaluation of the system at different CSI sounding rates, its effectiveness is demonstrated in the range of 30 Hz to 1500 Hz. Furthermore, the computational resources required for real-time operation of WI-MOID are evaluated, and it is shown that WI-MOID can operate in real time on commercially available edge devices.

[0254] Moving objects can be classified as human or non-human, the latter including a variety of entities such as pets and robot vacuum cleaners. In the following description, human motion refers to the motion of a human body, such as walking, running, sneaking, etc. In contrast, non-human motion refers to the motion of animate or inanimate objects, including animals, robots, and other physical objects. In some embodiments, the system classifies any motion originating from a non-human subject as non-human motion.

[0255] Interestingly, pets, cleaning robots, and even electrical appliances like fans can generate motions that can be picked up by sensing systems, but these motions differ from those of humans. Differences in gait patterns are a key aspect of this difference, with humans, pets, and robots each exhibiting unique motion patterns due to their bipedal, quadrupedal, and wheeled movements. These differences are visually represented in Figure 11. Due to these different motion patterns, analyzing features such as speed patterns becomes essential to distinguish between the motions of these objects.

[0256] The schematic diagram of the system WI-MOID 1200 shown in FIG. 12 is mainly composed of the following three modules.

[0257] First, the system 1200 includes a data processing module 1210 that preprocesses the CSI received by the receiver. The data processing module 1210 calculates the ACF of the CSI. Based on the ACF, the data processing module 1210 calculates motion statistics to determine the presence of a moving target in the environment. Furthermore, the data processing module 1210 can derive the velocity of the moving target based on the ACF.

[0258] Second, the system 1200 comprises a feature extraction module 1220, which effectively extracts a set of features from a moving target. Using the ACF, motion statistics, and velocity, the feature extraction module 1220 can extract physically interpretable features and signal statistical characteristics of the moving target. These features are independent of the environment, position, and orientation.

[0259] Last but not least, the system 1200 includes a classification module 1230 that uses an SVM and a state machine to output a final identification result of the detected motion. In some embodiments, the classification module 1230 first classifies the target features using an SVM, then employs a state machine to adjust for any potential misclassification based on the transition characteristics between states, and then finally outputs the result.

[0260] The present teachings disclose a novel method for detecting object motion and estimating its velocity through walls using a single WiFi link.

[0261] In wireless communication systems using omnidirectional antennas, WiFi signals propagate through indoor environments and experience multiple reflections due to the presence of various reflective surfaces. The CSI of a WiFi signal provides information about the wireless channel between the transmitter and receiver. The CSI of a multipath channel at a given time t and frequency f can be expressed as: TIFF2025066624000014.tif14158where a l (t) and τ l (t) respectively represent the complex amplitude and propagation delay of the l-th multipath component (MPC), and L represents the number of MPCs contributing to the CSI.

[0262] To accurately detect target motion in indoor environments, it is crucial to consider the presence of numerous multipath signals arising from abundant scattering. By employing a statistical model that takes into account all these multipath components, the target's velocity can be estimated.

[0263] Indoor environments experience abundant scattering from both static and dynamic scatterers. Static scatterers include walls, floors, and stationary furniture, while dynamic scatterers include moving individuals or objects. The principle of electromagnetic (EM) superposition allows the CSI to be decomposed as follows: TIFF2025066624000015.tif12158 where Ω s (t) represents the set of static scatterers, and Ω d (t) corresponds to a set of dynamic scatterers, and H i (t,f) and H j (t,f) represent the contribution of the ith and jth scatterers, respectively. The noise term ε(t,f) is i (t,f) and H j (t,f) are statistically independent. Each scatterer acts as a "virtual transmitter" and scatters the received EM waves into its surroundings. CSI represents the aggregate electric field of all received EM waves.

[0264] ACF(ρ H The covariance between H(t,f) and H(t+τ,f) can be calculated by dividing the covariance between H(t,f) and H(t+τ,f) by the variance of H(t,f) itself. Mathematically, this is expressed as: TIFF2025066624000016.tif30158Here, J0(·) is the Bessel function of the first kind, given by the following equation. TIFF2025066624000017.tif12148δ(-) represents the Dirac delta function.

[0265] The operating statistics φ(f) for a subcarrier of frequency f are defined as the ACF of the CSI H(t,f) with a time lag τ=1 / F s , F s is the sounding rate, i.e., TIFF2025066624000018.tif16158

[0266] Motion statistics serve as a reliable measure of the presence or absence of motion in a given environment. In a static environment, the motion statistics φ(f) are close to 0, but in a dynamic environment with motion, φ(f)>0. Figures 13A-13C show the motion statistics for human, pet, and robot motion, respectively.

[0267] Despite the size discrepancy, various objects can cause similar confusion over different distances. This is evident from the similar motion statistics shown in Figures 13A-13C and 14A. As a result, relying solely on motion statistics cannot yield reliable recognition results. Considering that speed can effectively capture distinct walking patterns exhibited by human and non-human subjects, we can extend the feature set by extracting speed information from WiFi signals.

[0268] To approximate the target motion speed, the ACF of CSI can be simplified by adopting a motion model that assumes the torso contributes most of the strong scatterers, resulting in ρ H (τ,f)=α(f)·J0(kvτ), where α(f) represents the gain of each subcarrier. This function is similar to the well-known Bessel function J0(x), where x=kvτ. By matching the location of the first peak or valley of J0(kvτ) with that of the Bessel function, the system can estimate the velocity of the object. Specifically, the system detects the first peak and calculates the velocity as TIFF2025066624000019.tif18148, where x0 represents a constant value corresponding to the first peak of J0(x), TIFF2025066624000020.tif12148 is the time lag corresponding to the first peak of J0(kvτ). The estimated velocities of human, pet, and robot motion obtained from the ACF are plotted in Figures 13A-13C, respectively.

[0269] Estimating the velocity of moving objects provides important insights for object identification. However, relying solely on velocity values ​​to distinguish between human and non-human objects is inappropriate because velocity values ​​can overlap, as shown in Figures 14A-14B. This is because object motions can exhibit significant overlap in velocity values. Therefore, to successfully identify moving objects, it is essential to comprehensively analyze the motion statistics, velocity, and ACF patterns over time. By extracting physically and statistically explainable features from these patterns, the system can strengthen discrimination and accurately distinguish between human and non-human objects. The next section provides a detailed explanation of these features and their role in motion classification.

[0270] The present teachings also disclose a methodology for distinguishing between the actions of humans, pets, and cleaning robots. In some embodiments, a set of features is extracted, and then a recognition model is developed that utilizes the selected features.

[0271] A. Feature Extraction

[0272] To identify features that effectively distinguish between human and non-human subject motion, a two-pronged approach can be adopted: examining physical motion signatures and statistical properties derived from CSI. On the physical side, we can delve into the gait-related attributes and speed patterns exhibited by subjects. On the statistical side, we can scrutinize the statistical signatures of ACFs and changes in motion statistics. This combined approach aims to encapsulate the holistic nature of motion and its impact on WiFi signals, facilitating accurate identification across a variety of subjects.

[0273] 1) Physical Feature Extraction: "Gait" refers to the pattern of limb movement when moving across a solid surface, and has proven to be a useful feature for human identification. Humans walk on two legs, pets such as dogs and cats primarily walk on four legs, and cleaning robots have wheel-based movements. These inherent differences result in unique gait and speed patterns, allowing us to distinguish between humans, pets, and robots.

[0274] When moving, the movement patterns of an animal's limbs create a gait, and humans and pets move their feet alternately. The stride cycle consists of a stance phase (when the foot touches the ground) and a swing phase (when the foot is raised and moves forward). The change in speed is the integral of the gait, and as shown in Figure 15, it decreases during the stance phase and increases during the swing phase. In contrast to humans and pets, cleaning robots do not exhibit large speed fluctuations, making them easy to identify through walking detection. In this system, the movement speed is estimated using the above method, and the speed pattern can be identified by analyzing peaks and valleys. The presence or absence of walking was determined by detecting speed fluctuations with clear ups and downs.

[0275] Stride cycle time and stride length. Stride cycle time and stride length are essential indicators of walking. Stride cycle time refers to the time it takes for the same foot to touch the ground twice, and stride length refers to the distance traveled by a person or animal between two consecutive steps. These definitions also apply to four-legged pets such as dogs and cats. Pets generally have shorter legs than humans, so their stride length is smaller than humans when moving around indoors.

[0276] This system extracts these two features from the velocity curves and can distinguish between human and pet motion. Specifically, stride cycle time is extracted by calculating the duration of three consecutive velocity peaks. Stride length is the integral of velocity values ​​over stride cycle time. Figures 16A and 16B show the average stride lengths of humans and dogs across 10 motion instances. It can be observed that although the stride cycle times of humans and pets are similar, their stride lengths are significantly different. Due to the short legs of pets, their stride lengths are significantly shorter than humans.

[0277] Speed ​​Average and Deviation. Due to the lack of feet on a wheeled platform, a vacuum cleaner robot typically maintains a constant speed while moving from one location to another and slows down when encountering an obstacle. Therefore, speed-based features are analyzed to distinguish between the vacuum cleaner robot and human behavior. The average speed is calculated as the average of the instantaneous speeds of the moving target in a specific time window. Furthermore, as depicted in Figures 17A-17B, speed deviation is considered to distinguish between human, pet, and robot behavior. Specifically, the system can estimate the variance, 25th percentile, and 75th percentile of the speed within the time window as speed deviation features, as shown in Figures 17A-17B.

[0278] Because vacuum cleaners have wheels and no legs, they maintain a relatively constant speed and only slow down when they encounter an obstacle. Therefore, speed patterns can be used to distinguish between vacuum cleaner robots and human behavior. Specifically, we extract the mean speed and speed deviation. The mean speed is derived from the average of the target's instantaneous speed within a specific time window. For the speed deviation, we can estimate the variance, 25th percentile, and 75th percentile, as shown in Figures 17A-17B.

[0279] 2) Statistical Feature Extraction: In most cases, human and pet behaviors exhibit characteristic gait features. However, certain behaviors, such as drinking, eating, and tail wagging, do not exhibit gait features. Therefore, to enhance the discrimination ability, the system further utilizes other features, such as statistical properties derived from ACF and CSI motion statistics. These properties are useful in situations where accurate speed estimation is unreliable, such as when the object is too far away for accurate gait estimation.

[0280] ACF Features. CSI's ACF serves as an advantageous statistic because it encompasses a wealth of information about the subject's behavior while exhibiting minimal sensitivity to location and environment. Therefore, the system extracts features based on the ACF or a function of the ACF. Several observations reveal that human, pet, and robot behaviors have unique impacts on the amplitude and frequency of the ACF peaks at each time step. This difference is evidenced by the variance in the amplitude and time lag of the ACF peaks and valleys for different types of behavior, as depicted in Figure 18A. To exploit these distinctions, the system can extract the mean values ​​of the ACF peaks and valleys separately for each type of behavior, as shown in Figures 18B and 18C. These features comprehensively represent the unique characteristics of human, pet, and robot behaviors, enabling effective discrimination between different behavior types in various environments.

[0281] Behavior Statistical Features. In addition to ACF-based features, behavior statistics also provide valuable insight into the statistical characteristics of the impact of a target's behavior on CSI. This helps improve the system's ability to distinguish between human and non-human behavior. To capture these statistical features, the system can calculate behavior statistics according to Equation (15) and extract their mean and variance as additional features. Statistical features enhance the system's resistance to environmental changes and enhance its adaptability to real-world scenarios.

[0282] B. Designing the recognition model

[0283] SVM is a supervised learning model that utilizes associated learning algorithms to classify data and perform regression analysis. In some embodiments, the disclosed SVM model uses a feature vector consisting of 11 features, including gait presence, stride length, stride cycle time, mean velocity, velocity variance, 25th percentile velocity, 75th percentile velocity, ACF peak mean, ACF valley mean, motion statistic mean, and motion statistic variance. During the training phase, SVM models employing linear or Gaussian kernels were trained, and the optimal model was selected using grid search complemented by five-fold cross-validation. Feature standardization was performed simultaneously. The trained SVM model was then tested in an unknown environment without further training or parameter tuning.

[0284] C. State Machine

[0285] In some cases, clear continuous gait features may not be obtained due to irregular movements such as stopping and turning around while walking, running, or moving slowly. As a result, the absence of clear gait features or the presence of small movements may lead the classifier to misclassify humans as non-humans. To address this challenge, our system utilizes a state machine based on HMM.

[0286] The disclosed state machine uses maximum likelihood estimation to estimate parameters and determine the most likely probability of an observation sequence occurring, where the emission probability b j Let (k) denote the probability of observing symbol k in state j, TIFF2025066624000021.tif18148 where c j The transition probability a can be estimated as follows: (k) represents the number of times symbol k is observed in state j, and M represents the total number of symbols. i , j represents the probability of transitioning from state i to state j, TIFF2025066624000022.tif19148 where c i , jis the number of transitions from state i to state j, and N is the total number of states.

[0287] By utilizing hidden Markov state machines, the system can improve the performance of the disclosed classifier and minimize the occurrence of misclassification due to irregular human behavior.

[0288] Several evaluation methodologies, including experimental setup, data collection, and metrics, can be used to evaluate the presented system, WI-MOID. In some embodiments, to evaluate the disclosed system, WI-MOID is implemented using commercially available WiFi devices and experiments are conducted in four representative indoor environments. Specifically, Scenario I corresponds to a 76.96 m² apartment, Scenario II corresponds to a 70.64 m² townhouse, Scenario III corresponds to a 40.27 m² detached house, and Scenario IV corresponds to a 200.75 m² office building. To acquire data from different locations and heights, one transmitter and two to three receivers can be used for each data collection. The receivers were installed 4 to 10 m from the transmitter at various heights between 0.3 m and 1 m, taking into account the presence of walls. To improve efficiency and diversity, multiple transceiver pairs are used in each session, but note that the CSI data from different pairs is not combined. Feature extraction and recognition processing are performed based on a single WiFi transceiver pair. The CSI data was collected on a 5.8GHz channel with a bandwidth of 40MHz and a sounding rate of 1500Hz.,In each scenario, the movement areas of humans, pets, and cleaning robots are,highlighted.

[0289] In some embodiments, behavioral data can be collected from two human subjects, one male and one female, measuring 170 cm and 187 cm tall, a dog approximately 1 m long, and an iRobot i3 vacuum cleaner. All data collection was conducted in a natural environment without any restrictions. Specifically, human behavioral data was acquired in an environment where subjects were free to walk around and stop at will. They engaged in a variety of activities, including using their mobile phones and playing games. Data on the dog's behavior was collected continuously throughout the day, including walking around, eating, drinking, and wagging its tail. Behavioral data was collected while the vacuum cleaner was cleaning a floor containing furniture such as tables and chairs, and the robot was free to move within accessible spaces to perform its tasks.

[0290] In some embodiments, experiments can be conducted over a six-month period, with subjects initiating movements from any location. The duration of movements ranged from two seconds to several minutes during each data collection session. In total, approximately 287 minutes of human movement data, 185 minutes of pet movement data, and 100 minutes of robot movement data can be accumulated. Movement statistics were applied to detect movements and segment the data into six-second instances. This resulted in 2,318 instances of human movement data, 493 instances of pet movement data, and 796 instances of robot movement data. During the training and validation process, 90% of the data from scenarios I and II can be randomly assigned to the training dataset, and the remaining 10% can be reserved for validation. Data from scenarios III and IV can be used solely to test the model's resistance to unknown environments and can be excluded from the training and validation phases.

[0291] To evaluate the model's adaptability to various types of human behavior, we collected 356 instances of sneaking and 367 instances of running in scenarios II and III. Furthermore, to evaluate the model's versatility in recognizing various non-human object behaviors, we collected 306 instances of ceiling fan behavior in scenarios I and III. The diverse human and fan behavior data collected in scenarios I and II were incorporated into the training and validation datasets. Conversely, the data in scenario III was utilized to evaluate the model's performance in unknown environments, comprehensively evaluating the model's generalization and adaptability.

[0292] The effectiveness of WI-MOID is evaluated by the following key metrics: The first metric is recall or TPR (True Positive Rate), which represents the likelihood that the system will correctly identify the target. The second metric is precision, also known as positive predictive value (PPV), which indicates the proportion of correct recognition results for a particular class. The third metric is "precision," which is the proportion of correctly identified data relative to the total data pool. The fourth metric is false alarm or false positive rate (FPR), which measures the probability that the system will misidentify the target's behavior.

[0293] A system's performance is considered improved when its precision, accuracy, and recall increase, along with a decrease in the false alarm rate. These metrics collectively serve as measures of the system's effectiveness.

[0294] To evaluate the effectiveness of WI-MOID, experiments were conducted using data collected from Scenarios I and II as training and validation data sets. First, features were extracted from the data and an SVM model trained with different kernels. To ensure the model's generalization ability, a five-fold cross-validation strategy was incorporated to randomize the training and validation data sets. The model with the highest validation accuracy was selected as the final model. The performance of the optimal SVM model was evaluated by its accuracy, precision, recall, and false alarm rate. As a result, WI-MOID achieved a high validation accuracy of 97.34% and a minimal false alarm rate of 1.75%, demonstrating its ability to accurately distinguish between human and non-human subject motions.

[0295] To evaluate the generalizability of the trained model across new contexts, we employ a test dataset collected from previously unobserved settings (Scenarios III and IV). The model's test accuracy was confirmed without further retraining. The results show that the pre-trained model achieved high test accuracy and precision of 92.61% and 93.93%, respectively, along with a low false alarm rate of 3.66%. These results demonstrate the robustness of WI-MOID in recognizing humans, pets, and robots through walls in novel environments. Furthermore, we confirmed that the trained model is robust to environmental changes and independent of the training environment. This property facilitates fast and easy deployment to new contexts, eliminating the need for additional user effort.

[0296] While pets and robots are typically the primary sources of non-human activity in indoor scenarios, electrical appliances such as electric fans can also contribute to this behavior. Therefore, it is necessary to evaluate the robustness of the disclosed algorithm in recognizing behavior caused by electrical appliances. In particular, we can evaluate the robustness of WI-MOID to the behavior of electric fans. To do so, we collect 306 instances of CSI data in an environment where an electric fan is operating. The results of validation accuracy, test accuracy, and false alarm rate show that the validation accuracy and test accuracy are high and the false alarm rate is low, even when fan activity is included. These results demonstrate that the disclosed features are effective and the disclosed model can be applied to the behavior of other electrical appliances, such as electric fans. The disclosed approach encompasses the extraction of physically meaningful attributes while simultaneously encapsulating statistical elements. These elements effectively reflect the impact of human and non-human behavior on CSI in the environment. Because electrical appliances generally operate continuously and cyclically, their impact on statistics differs from human behavior. Therefore, the disclosed model can robustly distinguish electrical appliance activity from human activity.

[0297] To measure the flexibility of the disclosed model in processing other types of human motion, sneaking and running activities were incorporated into the experimental setup. A total of 106 minutes of CSI data encompassing these motions was collected, and 723 6-second instances representing various human motions were extracted. The system trained and validated the SVM model using various human motion data from Scenarios I and II. The model's robustness was further evaluated by testing it with data from Scenarios III and IV, which were not included in the initial training set. All data subsets used for training, validation, and testing consisted of a comprehensive array of human motions, including walking, running, and sneaking. As a result, a high recognition accuracy of 92.55% was achieved, even when including a wider variety of human motions. This highlights the adaptability of WI-MOID to various human motions.

[0298] To evaluate the effectiveness of the disclosed state machine, we compare the recognition accuracy of WI-MOID with and without the state machine. The results of the recognition accuracy and false positive rate for humans, robots, and pets show that using the state machine improves the recognition accuracy of humans, robots, and pets by 1.67%, 1.85%, and 17.98%, respectively, compared to without the state machine. The state machine significantly improves pet recognition accuracy by addressing challenges posed by pet behavior that are difficult for SVM classifiers to handle. Specifically, pets can be misclassified as humans due to similar gait patterns. Furthermore, pets can be mistaken for cleaning robots when they generate small movements, such as moving slowly or wagging their tails, that interfere with the CSI signal in a similar manner to cleaning robots. Furthermore, the disclosed state machine reduces the false positive rate for human, robot, and pet recognition by 1.16%, 2.56%, and 1.97%, respectively. The reduction in false positives and the improvement in human detection accuracy are particularly noteworthy because the state machine helps avoid misclassification when human behavior does not have a clear gait pattern. These results show that the state machine can effectively improve the recognition accuracy, reduce the false alarm rate, and enhance the robustness and reliability of WI-MOID.

[0299] To evaluate the effectiveness of the engineered features in the disclosed model, a comparative analysis can be performed using a model trained with a 95% principal component analysis (PCA)-driven approach designed to reduce the dimensionality of the feature vector. This strategy preserves only the principal components that account for 95% of the total variance. The SVM model was retrained with the same dataset but with a condensed feature set. The comparison showed that the accuracy of the model trained with PCA was lower than that of the original model utilizing the engineered features. This observation demonstrates the effectiveness of the engineered features, which contribute to the high accuracy of the disclosed model.

[0300] In some embodiments, we investigated the impact of motion segment length on WI-MOID performance. We found that motion segment length does not significantly impact system accuracy. While shorter motion segments may provide more specific information about motion, longer motion segments may provide a broader view of overall motion. However, the results suggest that neither length significantly impacts system accuracy. Therefore, WI-MOID is robust to changes in motion segment length from 6 seconds to 25 seconds and can accurately analyze motions of different lengths.

[0301] In some embodiments, the impact of different sounding frequencies on performance can be investigated. The detection rate and false alarm rate of WI-MOID can be compared at four different sounding frequencies: 1500 Hz, 500 Hz, 100 Hz, and 30 Hz. The results show that the disclosed algorithm achieves high accuracy and low false alarm rate across a range of sounding frequencies, with accuracy improving as the sounding frequency increases. Notably, the accuracy of the algorithm remains satisfactory even at low sounding rates, demonstrating robustness to variations in data acquisition frequency. This suggests that the algorithm can be effectively used without sacrificing accuracy in resource-constrained scenarios where high sounding rates are not feasible due to memory or power limitations. Therefore, the disclosed algorithm provides a scalable solution for real-time data processing applications on resource-constrained edge devices.

[0302] The computational efficiency and storage requirements of the disclosed algorithm were evaluated on a commercially available edge device equipped with a Marvell chipset. The results showed that the algorithm is highly optimized for edge computing, with minimal CPU and storage consumption, making it an ideal candidate for deployment on resource-constrained edge devices. Furthermore, the algorithm's scalability allows it to accommodate a variety of requirements and device specifications, making it a versatile solution for real-time data processing at the edge.

[0303] WiFi has become increasingly popular as a primary communication medium, becoming virtually ubiquitous. Channel state information (CSI) conveys information about the surrounding environment, allowing for "visibility" of objects in the environment. This has enabled many applications through WiFi sensing. In contrast to cameras and millimeter-wave radar, WiFi has a wide coverage area and can operate under both line-of-sight (LOS) and non-line-of-sight (NLOS) conditions, providing a superior level of privacy protection. However, the majority of WiFi sensing applications assume that all activity in the environment is due to human activity and do not fully consider the influence of non-human objects. This is due to the difficulty of distinguishing between humans and non-humans using WiFi, as both can cause similar variations in CSI. However, non-human elements such as pets, robot vacuum cleaners, and electric fans are ubiquitous in diverse environments. Therefore, it is essential to develop algorithms that can distinguish between human and non-human objects.

[0304] Current efforts to distinguish between humans and non-human objects using WiFi generally fall into two categories: traditional model-based and machine learning-based. Traditional model-based methods have high interpretability but require precise device placement and height, and their effectiveness decreases as the complexity of the environment increases. In contrast, machine learning-based methods utilize extracted meaningful features, which can alleviate limitations imposed by device placement and environmental constraints, thereby broadening their applicability. These methods, which rely on hand-crafted features, retain excellent interpretability similar to model-based methods. However, the performance of traditional machine learning models depends on the effectiveness of these hand-crafted features. If these features cannot be accurately derived from the data, their performance will degrade. Furthermore, models based on traditional machine learning classifiers, such as support vector machines (SVMs), tend to perform poorly on classification problems that exhibit significant overlap and struggle with linearly inseparable data.

[0305] In recent years, deep learning models have achieved remarkable success in areas such as image processing and speech processing, demonstrating their powerful ability to extract and classify features from highly complex and high-dimensional data. This has also sparked interest in the field of WiFi sensing, and many studies have begun to adopt deep learning. However, these studies have simply adopted specific existing network frameworks. While various types of deep learning models have excelled in fields such as image processing and language learning, current WiFi sensing research has not extensively compared the performance of various models on the task, resulting in insufficient investigation into the suitability of various models for WiFi sensing tasks. Furthermore, currently, most deep learning-based WiFi sensing tasks use raw CSI as input, which makes it impossible to mitigate the impact of environmental factors on feature extraction. As a result, trained networks often fail when transferred to new environments. Although various strategies for domain adaptation have been proposed, large amounts of user data are required to retrain or tune the model in new environments.

[0306] The present teachings disclose a framework that utilizes deep neural networks to recognize motion from human and non-human objects using WiFi signals. The disclosed method utilizes the amplified autocorrelation function (A-ACF) extracted from CSI as the network input, enabling the network to extract only features related to the object's motion, regardless of changes in the object's position, orientation, or environment. The disclosed system can utilize two major categories of deep learning models: convolution-based models derived from image processing and recurrent neural network-based models derived from natural language processing. The performance of these models in human and non-human recognition tasks and their computing resource requirements can be evaluated. Furthermore, the effectiveness of porting pre-trained models from the image processing field to WiFi sensing tasks is verified.

[0307] In some embodiments, the present teachings disclose an environmentally robust indoor object classification algorithm. By designing the amplified autocorrelation function (ACF) of channel state information (CSI) as the network input, the algorithm effectively extracts features of moving objects for identification. Because these features are independent of environmental conditions, position, and orientation, the model is insensitive to environmental fluctuations. Furthermore, the effectiveness of different neural networks is compared in identifying human and non-human objects in various environments using WiFi. The performance of the networks is extensively evaluated under known and unknown environments, including situations with unknown objects and situations with multiple objects simultaneously in unknown environments. The feasibility of transferring pre-trained models from image processing tasks to WiFi datasets is accordingly explored.

[0308] In some embodiments, WiFi signals emitted from a transmitter traverse the environment by reflecting multiple times off various surfaces, including people and objects, before reaching a receiver. WiFi-based sensing systems operate by analyzing the effect of these people and objects on the propagation of WiFi signals, thereby detecting their movements and changes. CSI is commonly employed as input for WiFi-based sensing systems because it encapsulates the direct propagation of a signal between a transmitter and a receiver and reflects the processes of reflection, scattering, and attenuation during this propagation. Due to multipath effects, CSI is commonly expressed as a collection of radio wave propagations along different paths: TIFF2025066624000023.tif14158Here α l is the multipath coefficient of the l-th component, and τ l is the associated propagation delay.

[0309] In practice, imperfect synchronization in commercial WiFi devices often introduces random noise in the CSI phase, which is difficult to efficiently remove, so researchers typically rely on the more reliable CSI amplitude, which can be measured via the power response. TIFF2025066624000024.tif24158Here, n(t,f) is the power density σ n 2 (f) is additive white Gaussian noise (AWGN), and |H(t,f)| 2 is statistically independent from

[0310] Since SI embodies the cumulative effect of the environment on signal propagation, it incorporates not only the effects of dynamic entities in the environment but also the effects of static objects such as walls, furniture, and floors on the signal. As a result, WiFi sensing systems that directly use CSI as input are significantly affected by the environment. They exhibit poor domain adaptation capabilities, requiring data collection and retraining in new environments. This hinders rapid deployment and use in new environments.

[0311] The extracted CSI differs between Environment A and Environment B for different moving objects, such as human motion, pet motion, and robot vacuum cleaner motion. The furniture in Environment B was moved between each data collection session. Based on the CSI observations, environmental factors dominate the CSI variability, significantly outweighing the influence of different moving objects, making it difficult to extract motion-related features. Therefore, eliminating the dependency of environmental context is crucial for WiFi sensing systems to extract essential statistics related to motion analysis and improve accuracy and deployment feasibility.

[0312] To separate the influence of the static environment and dynamic subjects on the signal, the system can first calculate the autocorrelation function (ACF) of the CSI, which encapsulates only the characteristics of the dynamic objects. Then, it calculates the robust statistical A-ACF (Amplified ACF), which amplifies the features of the dynamic objects and facilitates the analysis of their motion patterns.

[0313] Specifically, we can extract the ACF of G(t,f) as follows: TIFF2025066624000025.tif28158Here, τ is the time lag. σ S 2 (f) and ρ S (τ) are the variance and ACF of the propagated signal, respectively. The normalized channel gain at frequency f is Defined as TIFF2025066624000026.tif18148.

[0314] To better extract the features of moving objects, a Maximum Ratio Combining (MRC) approach can be adopted for ACF. This combining across all subcarriers improves the signal-to-noise ratio (SNR) of the ACF, thereby highlighting the effect of moving objects on the signal. The aggregated ACF after MRC is expressed as: TIFF2025066624000027.tif24158 where N s is the total number of subcarriers. As τ → 0, we have: TIFF2025066624000028.tif14158

[0315] The target action is continuous, TIFF2025066624000029.tif12148 Next TIFF2025066624000030.tif12148. Sounding rate F s If is high, w(f i )of TIFF2025066624000031.tif12148. Then, the aggregated ACF can be estimated by: TIFF2025066624000032.tif16158

[0316] In some embodiments, to amplify the velocity information, the difference of the aggregated ACFs is calculated using You can take TIFF2025066624000033.tif12148. We use Δρ(τ) to represent the A-ACF. It can be expressed as TIFF2025066624000035.tif13148.

[0317] Comparing the results of A-ACF extraction from CSI data with the original CSI data shows that the behavior characteristics of the same type of object in different environments are similar, while the A-ACF patterns of objects with different behaviors show large discrepancies, whether in the same or different environments.

[0318] Because the A-ACF encapsulates only dynamic features related to the object's motion and does not include any environmental or directional information related to the object, the disclosed WiFi sensing system utilizing the A-ACF as input is insensitive to variations in the environment and object's position. In some embodiments, the system ignores aspects such as the object's orientation, position, and surrounding environment, focusing instead on extracting more essential features that correlate with the object's motion. This independence allows the disclosed system to be quickly and efficiently deployed in new environments.

[0319] The system can then derive robust motion statistics from the A-ACF and use them to detect and segment motion-containing fragments. For a power response G(t,f), the subcarrier f i The robust performance statistics derived from the ACF at time t above are defined as follows: TIFF2025066624000036.tif12132Here F s is the sounding rate.

[0320] Robust motion statistics serve as a reliable measure of the presence or absence of motion in a given environment. In a stationary environment, the robust motion statistics φ G (f) is close to 0, but in a dynamic environment with motion, φ G (f)>0. Following this, the A-ACF spectrum can be divided into 5-second segments that are detected to contain motion, as shown in Figure 19. The extracted A-ACF spectrum has a size R N*T where N is the total number of time lags at each time instance and T is the number of time instances in the segment.

[0321] To respond to various model types, the system can further convert the 2D A-ACF spectrum into inputs according to each model, which mainly includes the following four cases:

[0322] Input of FNN. In a feedforward neural network (FNN), individual samples are usually represented as a flattened one-dimensional vector. This is because FNNs are designed to process each feature independently and do not consider the inherent structure or correlation between features. Therefore, the two-dimensional A-ACF spectrum R N*T is a one-dimensional vector R to meet the input requirements of the FNN. NT is flattened to

[0323] Input for image-based models. Unlike FNNs, image-based models like CNNs are specifically designed to handle multidimensional data like images while preserving the spatial relationships between pixels and features. Therefore, we can directly input the 2D A-ACF spectrum into image-based models.

[0324] Input for Language-Based Models. Language-based models, such as RNNs, are particularly well-suited to dealing with sequential data, where the order of the input matters. RNNs maintain hidden states that can, in theory, capture information about past elements in the sequence. Thus, the typical input for an RNN is a sequence. In this setup, we take a sequence of A-ACF vectors, R N are input to the network one at a time.

[0325] Input for Transformer-Based Models. Transformer-based models have been applied to both natural language processing (NLP) and vision tasks. In both tasks, transform models treat the data in a similar sequential manner. In a two-dimensional A-ACF spectrum, we have P small patches R H*W These patches are divided into a one-dimensional vector R HW A position embedding is added to the vector to provide information about the relative position of the patch in the original image.

[0326] Figure 20 provides an intuitive representation of how A-ACF data are integrated into these models.

[0327] To provide a concise overview of various deep learning models that have shown promising performance in the computer vision and natural language processing domains, we delve into their advantages and limitations when applied to human and non-human activity identification based on WiFi signals, and also provide a discussion of their applications in transfer learning and wireless sensing.

[0328] A forward neural network (FNN) is the simplest type of artificial neural network. In an FNN, information moves in only one direction: from the input layer through the hidden layer to the output layer. There are no loops in the network; information is always sent forward. Each neuron in a layer receives input from all neurons in the previous layer, processes the input using a weighted sum and bias, applies an activation function to this sum, and passes the result to all neurons in the next layer. A multilayer perceptron (MLP) is a feedforward neural network architecture widely used for machine learning tasks. An MLP takes a flattened A-ACF as input and maps latent features to a categorical space with nonlinearity introduced by the activation function. While MLPs can model complex nonlinear relationships in data, and their hidden layers extract increasingly abstract and meaningful features, MLPs do not explicitly consider spatial and temporal relationships in the data. This limitation makes MLPs unsuitable for modeling spatial and temporal dependencies between human and nonhuman behavior from A-ACFs, yet they are employed as classifiers in other networks, such as CNNs.

[0329] Convolutional neural networks (CNNs) are primarily used in the field of computer vision and have achieved state-of-the-art results in image classification, object detection, and many other tasks. CNNs are designed to automatically and adaptively learn a spatial hierarchy of features from input data. CNNs have one or more convolutional layers, often followed by a pooling layer, and finally a fully connected layer for classification or regression. The convolutional layers are designed to recognize small, simple patterns, while later layers recognize more complex structures. In CNNs, the A-ACF is input to the network as a two-dimensional image, and a two-dimensional convolution kernel with a sliding window mechanism is applied to estimate a feature map from the A-ACF spectrum. Using a two-dimensional CNN for feature extraction from the A-ACF spectrum of motion has many advantages. First, motion features are captured using a two-dimensional kernel that effectively identifies the structure and characteristics of the ACF spectrum from two dimensions. Furthermore, parameter sharing is employed to significantly reduce the number of parameters, improve computational efficiency, and mitigate the risk of overfitting. However, it is worth noting that CNNs primarily focus on local patterns and may miss distant dependencies and global context present in the data.

[0330] Recurrent neural networks (RNNs) are specifically designed to handle sequence data. Including loops in the network allows information to flow from one step in a sequence to the next. This allows RNNs to process data such as time series and sentences. However, standard RNNs often struggle to learn distant dependencies due to vanishing or exploding gradients. This has led to the development of more advanced types of RNNs, such as long short-term memories (LSTMs) and gated recurrent units (GRUs). In RNNs, A-ACFs are interpreted as time-series signals. The network treats each instance of A-ACF as a separate token. RNNs are well-suited to understanding time dependencies within A-ACF segments, making them an ideal choice for analyzing the time-varying features of A-ACFs induced by behavior. However, they are poor at recognizing the ACF features at specific moments, resulting in the neglect of A-ACF features resulting from short-term behavior.

[0331] Transformers are a type of model that uses a self-attention mechanism and has achieved state-of-the-art results in various natural language processing and computer vision tasks. Unlike RNNs, Transformers process input data in parallel rather than sequentially, making them more efficient. RNNs are designed to handle sequences and can capture complex patterns in data, making them suitable for tasks such as machine translation and text summarization. To better capture the two-dimensional features of A-ACF spectra, we employ a transformer-based vision model, the Vision Transformer (ViT). ViT processes the input A-ACF spectrum by dividing it into small, fixed-size image patches and reshaping them into vectors. These patch vectors are then linearly transformed into D-dimensional embeddings. To preserve location information, a location embedding is added to these patch embeddings. These patch embeddings are then passed to a Transformer encoder, which contains a multi-head self-attention mechanism and multiple layers of a location-wise fully connected feedforward network. Finally, the output corresponding to the first token is used by a classification head to generate the final prediction for human and non-human action classification.

[0332] In some embodiments, a CNN+RNN model combines the advantages of CNN and RNN. Typically, a CNN component is used to extract a sequence of meaningful features from input data (e.g., frames of a video), and then an RNN component is used to analyze the sequence of features. This combination makes the CNN+RNN model effective for tasks involving both spatial and temporal dependencies, such as video analysis and image captioning. For A-ACF input, the system can treat it as a spectral image, and the CNN network applies a 2D convolution kernel to extract spatial information from the spectrum. A GRU can then be applied to extract temporal information from the extracted feature map.

[0333] Transfer learning is a machine learning technique that uses a model pre-trained on a large dataset as a starting point for a desired task, often by leveraging knowledge learned in one problem domain (the source domain) to a different, related but different problem domain (the target domain). This is particularly useful when labeled data for the target task is limited. In deep learning, models trained on large image datasets are often used for other vision tasks.

[0334] In some embodiments, due to the scarcity of labeled WiFi data and the transformation of WiFi signals into two-dimensional A-ACF spectra, the feasibility of leveraging pre-trained image models in WiFi sensing tasks can be investigated. By utilizing A-ACF spectral images as input for a transfer learning network, the system can transfer features learned by a deep neural network from image data to WiFi-based human and non-human action identification tasks, aiming to enhance the performance of WiFi sensing systems. In some embodiments, the effectiveness of transferring a ResNet-18 model pre-trained on the ImageNet dataset to identify human and non-human actions based on WiFi signals is evaluated and compared with the performance of a ResNet-18 model explicitly trained on WiFi signals. Furthermore, the impact of fine-tuning different layers on transfer performance can be investigated.

[0335] In some embodiments, a framework is introduced to evaluate the disclosed system, including a pair of devices equipped with commercially available WiFi network interface cards. One device acts as a transmitter (Tx) and the other as a receiver (Rx). Each device is equipped with two omnidirectional antennas, for a total of four communication links per transceiver unit. These antenna pairs facilitate streaming of CSI across 58 subcarriers. The framework operates within WLAN channel 153, utilizing a carrier frequency of 5.18 GHz and a bandwidth of 40 MHz. The Tx actively deploys sounding frames at a sounding rate of 1500 Hz.

[0336] In some embodiments, the disclosed framework is evaluated in five representative environments with a total of 29 different configurations, including a compact residential apartment, a townhouse, two single-family homes, and an office building. Floorplans of these environments are shown, along with designated locations of the Tx (shown in blue) and Rx (highlighted in orange). Data can be collected for both line-of-sight and non-line-of-sight conditions to measure the system's through-wall capabilities. In some embodiments, the analysis employs single-link WiFi data only. The distance between the Tx and Rx ranges from 2 to 8 meters.

[0337] In some embodiments, CSI data can be collected for four types of objects (humans, pets, vacuum robots, and fans) across five scenarios (Scenarios 1–5). Some humans and pets appear in multiple scenarios, while others appear in only one scenario. The evaluation metric employed to measure the accuracy of the model is top-1 accuracy, defined as the percentage of correctly classified motion segments from the dataset. CPU prediction time and FLOPS are used to evaluate computational requirements and complexity. Peak memory usage, parameter size, and model size are evaluated to measure memory usage. Evaluating computational and memory requirements is crucial for deploying deep learning-based methods for WiFi sensing on edge devices with limited computational resources.

[0338] In some embodiments, five experiments based on different environmental settings allow for thorough evaluation of various deep learning models. In each experiment, data from four scenarios can be specified for training and validation, and one other scenario can be specified for testing. The training and validation datasets were randomly split in an 8:2 ratio. The accuracy of different models is compared on an unknown validation dataset. The MLP model showed stable performance with an average accuracy of 91.59%. LeNet performed slightly better with an average accuracy of 93.60%. Models from the ResNet family performed well, with ResNet101 achieving the highest average accuracy of 96.38%. Other deep learning models, such as RNN, GRUNet, and LSTM, performed relatively poorly, with average accuracies of 86.64%, 89.77%, and 85.79%, respectively. The ViT model also achieved a commendable average accuracy of 92.40%. Overall, the results convincingly demonstrate the superiority of the disclosed deep learning-based framework in accurately identifying human and non-human objects through walls using WiFi. The remarkable performance of the ResNet model in particular confirms the effectiveness of deep learning approaches in complex detection tasks.

[0339] To evaluate the model's performance in unknown environments, it can be tested from three different perspectives: recognizing known objects in unknown scenarios, identifying unknown objects in unknown scenarios, and dealing with settings where multiple objects coexist.

[0340] Recognizing Unknown Objects in Unknown Environments. We first evaluate the model's ability to accurately identify known objects in unknown environments. Following a leave-one-out approach, we conducted five testing experiments, and ResNet18 achieved an impressive average test accuracy of 91.71%. All ResNet models exceeded the 90% mark in average test accuracy. These results demonstrate that the disclosed architecture minimizes the impact of factors such as environment, position, and orientation on recognition performance. This enables effective discrimination between humans and non-humans in novel environments without the need for additional training or parameter tuning.

[0341] Recognizing Unknown Objects in Unknown Environments. Evaluation explores the model's robustness to unfamiliar objects in untested environments. The evaluation uses the same leave-one-out training approach, where the objects embedded in the test data are not present in either the training or validation datasets. Surprisingly, even when encountering novel environments and objects with unfamiliar behaviors, the disclosed framework maintains high accuracy in identifying human and non-human objects. This robustness is due to the disclosed A-ACF-based framework's effective extraction of features related to behavior patterns, thereby capturing universal features even among subtle differences in the behavior patterns of different objects. This robust recognition capability supports the model's ability to be rapidly deployed in new environments with minimal user intervention.

[0342] Recognition in Unknown Environments with Multiple Coexisting Objects. Furthermore, we can verify the robustness of the disclosed framework when tasked with detecting multiple coexisting objects in unknown environments. For this evaluation, we collect data from two scenarios, Scenario 1 and Scenario 2, in which two individuals and two dogs move simultaneously. By applying a similar leave-one-out approach, we can evaluate the model's ability to recognize multiple objects in unknown environments. As a result, the ViT model achieved a test accuracy of 82.86%. This performance means that the disclosed model can accurately identify multiple objects, whether humans or pets, even when they are present simultaneously, eliminating the need for additional retraining or fine-tuning. This robustness further highlights the practical value and adaptability of the disclosed model in real-world scenarios.

[0343] Transfer Learning Performance. To investigate whether transfer learning can enhance WiFi-based human and non-human target recognition by leveraging features learned from image recognition, we evaluate the accuracy of a ResNet18 model pre-trained on ImageNet in five experiments. This is compared to a base ResNet18 model trained only on WiFi data. Furthermore, we investigate the impact of freezing the model for various epochs on the transfer learning results. Results suggest that setting the number of frozen epochs to 10 and 50 slightly improves accuracy, increasing by 0.65% and 0.48%, respectively. In some embodiments, indefinite model freezing resulted in a decrease in accuracy. This negative transfer is attributed to the distributional differences between the target dataset (WiFi data) and the source dataset (image data).

[0344] FIG. 21 shows a flowchart of an example method 2100 for wireless human-non-human motion detection according to some embodiments of the present disclosure. In various embodiments, method 2100 may be performed by the systems disclosed above. At operation 2102, a wireless signal is transmitted over a wireless multipath channel within a venue. At operation 2104, the wireless signal is received over the wireless multipath channel. The received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel being affected by the motion of an object within the venue. At operation 2106, time-series channel information (TSCI) of the wireless multipath channel is obtained based on the received wireless signal. At operation 2108, an autocorrelation function (ACF) is calculated based on the TSCI. At least one ACF feature, such as the ACF or a function of the ACF (e.g., a squared, logarithmic, or monotonic function of the ACF), is calculated at operation 2110. At operation 2112, ACF statistics, motion statistics, speed statistics, and gait statistics are calculated based on at least one ACF feature of the ACF or a function of the ACF. At operation 2114, in response to determining that motion is detected, an object associated with the detected motion is classified as either human or non-human based on the ACF statistics, motion statistics, speed statistics, and gait statistics.

[0345] FIG. 22 is a flowchart illustrating detailed operations 2200 for wireless human-non-human motion detection according to some embodiments of the present disclosure. In some embodiments, operation 2200 may be performed as part of operation 2112 of FIG. 21 . In operation 2210, ACF statistics are calculated based on a first ACF feature or function of the ACF. In operation 2220, motion statistics associated with object motion are calculated based on a second ACF feature or function of the ACF. In operation 2230, speed statistics associated with the speed of the object motion are calculated based on a third ACF feature or function of the ACF. In operation 2240, gait statistics associated with the object's walking motion are calculated based on a fourth ACF feature or function of the ACF.

[0346] FIG. 23 shows a flowchart of an example method 2300 for wireless sensing in accordance with some embodiments of the present disclosure. In various embodiments, method 2300 may be performed by the systems disclosed above. At operation 2302, a wireless signal is transmitted through a wireless multipath channel within a venue, where the wireless multipath channel is affected by the motion of an object within the venue. At operation 2304, the wireless signal is received through the wireless multipath channel. The received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel and the motion of the object. At operation 2306, time-series channel information (TSCI) of the wireless multipath channel is obtained based on the received wireless signal. At operation 2308, multiple autocorrelation functions are calculated based on the TSCI, and each autocorrelation function (ACF) is calculated based on the CI of the TSCI in a respective sliding time window. At operation 2310, at least one ACF vector is constructed. Each ACF vector is a vector associated with a respective ACF, including multiple vector elements each associated with a respective time lag. Each vector element is a respective ACF value evaluated at a respective time lag. In operation 2312, at least one ACF vector is rearranged into rearranged ACF data. Each ACF vector is a one-dimensional (1D) ACF block. In operation 2314, a wireless sensing task is performed based on a task engine that processes the rearranged ACF data as input.

[0347] The order of the actions in the figures may be changed according to various embodiments of the present teachings.

[0348] 24 is an exemplary block diagram of a first wireless device, e.g., a bot 2400, of a wireless detection, sensing, and / or learning system according to some embodiments of the present disclosure. The bot 2400 is an example of a device that can be configured to implement various methods described herein. As shown in FIG. 24, the bot 2400 includes a housing 2440 that includes a processor 2402, a memory 2404, a transceiver 2410 consisting of a transmitter 2412 and a receiver 2414, a synchronization controller 2406, a power module 2408, an optional carrier configurator 2420, and a wireless signal generator 2422.

[0349] In this embodiment, processor 2402 controls the general operation of bot 2400 and may include one or more processing circuits or modules, such as a central processing unit (CPU), and / or any combination of general-purpose microprocessors, microcontrollers, digital signal processors (DSPs), field programmable gate arrays (FPGAs), programmable logic devices (PLDs), controllers, state machines, gate logic, discrete hardware components, dedicated hardware finite state machines, or any other suitable circuits, devices, and / or structures capable of performing calculations or other manipulations on data.

[0350] The memory 2404 may include both read-only memory (ROM) and random access memory (RAM) and may provide instructions and data to the processor 2402. A portion of the memory 2404 may also include non-volatile random access memory (NVRAM). The processor 2402 typically performs logical and arithmetic operations based on program instructions stored in the memory 2404. The instructions (also known as software) stored in the memory 2404 may be executed by the processor 2402 to perform the methods described herein. The processor 2402 and the memory 2404 together form a processing system that stores and executes software. As used herein, "software" refers to any type of instructions, whether referred to as software, firmware, middleware, microcode, or the like, that can configure a machine or device to perform one or more desired functions or processes. The instructions may include code (e.g., code in source code format, binary code format, executable code format, or any other suitable format). The instructions, when executed by one or more processors, cause the processing system to perform various functions described herein.

[0351] The transceiver 2410, including a transmitter 2412 and a receiver 2414, allows the bot 2400 to send and receive data to and from a remote device (e.g., the Origin or another bot). The antenna 2450 is typically attached to the housing 2440 and electrically coupled to the transceiver 2410. In various embodiments, the bot 2400 includes multiple transmitters, multiple receivers, and multiple transceivers (not shown). In one embodiment, the antenna 2450 is replaced with a multi-antenna array 2450 capable of forming multiple beams, each pointing in a separate direction. The transmitter 2412 may be configured to wirelessly transmit signals having different types or functions, and such signals are generated by the processor 2402. Similarly, the receiver 2414 is configured to receive wireless signals having different types or functions, and the processor 2402 is configured to process multiple different types of signals.

[0352] The bot 2400 in this example can function as a bot or Type 1 device or sensing transmitter for wireless detection, sensing, and / or learning. The synchronization controller 2406 in this example is configured to control the operation of the bot 2400 to synchronize or asynchronously with another device, such as an Origin or another bot. In one embodiment, the synchronization controller 2406 can control the bot 2400 to synchronize with an Origin that receives a wireless signal transmitted by the bot 2400. In another embodiment, the synchronization controller 2406 can control the bot 2400 to transmit a wireless signal asynchronously with the other bots. In another embodiment, the bot 2400 and the other bots can each transmit a wireless signal individually and asynchronously.

[0353] The carrier configurator 2420 is an optional component within the bot 2400 for configuring transmission resources, e.g., time and carrier, for transmitting the wireless signal generated by the wireless signal generator 2422. In one embodiment, each CI in the time series of CIs has one or more components, each corresponding to a carrier or subcarrier of the transmission of the wireless signal. Wireless sounding sensing may be based on any one or more combinations of the components.

[0354] The power module 2408 can include a power source, such as one or more batteries, and a power regulator to provide regulated power to each of the above-mentioned modules in Figure 24. In some embodiments, if the bot 2400 is coupled to a dedicated external power source (e.g., a wall outlet), the power module 2408 can include a transformer and a power regulator.

[0355] The various modules described above are coupled together by a bus system 2430. The bus system 2430 may include, for example, a power bus, a control signal bus, and / or a status signal bus in addition to a data bus. It will be appreciated that the modules of the bot 2400 may be operably coupled to each other using any suitable technology and medium.

[0356] 24 illustrates a number of separate modules or components, those skilled in the art will appreciate that one or more of the modules may be combined or co-implemented. For example, processor 2402 may implement the functionality described above with respect to processor 2402 as well as the functionality described above with respect to wireless signal generator 2422. Conversely, each of the modules illustrated in FIG. 24 may be implemented using multiple separate components or elements.

[0357] FIG. 25 is an exemplary block diagram of a second wireless device, e.g., Origin 2500, of a system for wireless detection, sensing, and / or learning, in accordance with one embodiment of the present teachings. Origin 2500 is an example of a device that can be configured to implement various methods described herein. In this example, Origin 2500 functions as an Origin or Type 2 device, or a sensing receiver for wireless detection, sensing, and / or learning. As shown in FIG. 25, Origin 2500 includes a housing 2540 that includes a processor 2502, a memory 2504, a transceiver 2510 consisting of a transmitter 2512 and a receiver 2514, a power module 2508, a synchronization controller 2506, a channel information extractor 2520, and an optional motion detector 2522.

[0358] In this embodiment, the processor 2502, memory 2504, transceiver 2510, and power module 2508 operate similarly to the processor 2402, memory 2404, transceiver 2410, and power module 2408 of the bot 2400. The antenna 2550 or multi-antenna array 2550 is typically mounted in the housing 2540 and electrically coupled to the transceiver 2510.

[0359] The Origin 2500 may be a second wireless device having a different type from the first wireless device (e.g., the Bot 2400). In particular, the channel information extractor 2520 of the Origin 2500 is configured to receive a wireless signal through a wireless channel and obtain time-series channel information (CI) of the wireless channel based on the wireless signal. The channel information extractor 2520 can transmit the extracted CI to an optional motion detector 2522 or a motion detector external to the Origin 2500 to perform wireless sounding sensing within the venue.

[0360] The motion detector 2522 is an optional component of the Origin 2500. In one embodiment, it is located within the Origin 2500, as shown in FIG. 25. In another embodiment, it is located outside the Origin 2500, in another device such as a bot, another Origin, a cloud server, a fog server, a local server, or an edge server. The optional motion detector 2522 may be configured to detect sounding information from a vibrating object or source within the venue based on the motion information. The motion information may be calculated based on a time series of CIs by the motion detector 2522 or another motion detector outside the Origin 2500.

[0361] In this example, the synchronization controller 2506 is configured to control the operation of the Origin 2500 to be synchronized or asynchronous with other devices (e.g., bots, other Origins, or independent motion detectors). In one embodiment, the synchronization controller 2506 controls the Origin 2500 to synchronize with a bot transmitting a wireless signal. In another embodiment, the synchronization controller 2506 controls the Origin 2500 to receive wireless signals asynchronously with other Origins. In another embodiment, the Origin 2500 and other Origins may each receive wireless signals individually and asynchronously. In one embodiment, the optional motion detector 2522 or a motion detector external to the Origin 2500 is configured to asynchronously calculate respective heterogeneous motion information based on respective time series of the CIs.

[0362] The various modules described above are coupled together by a bus system 2530. The bus system 2530 may include, for example, a power bus, a control signal bus, and / or a status signal bus in addition to a data bus. It will be appreciated that the modules of the Origin 2500 may be operably coupled to each other using any suitable technology and medium.

[0363] 25 illustrates a number of separate modules or components, those skilled in the art will appreciate that one or more of the modules may be combined or co-implemented. For example, processor 2502 may implement the functionality described above with respect to processor 2502 as well as the functionality described above with respect to channel information extractor 2520. Conversely, each module illustrated in FIG. 25 may be implemented using multiple separate components or elements.

[0364] In some embodiments, a method for recognizing humans and non-humans is disclosed for distinguishing between human and non-human actions (e.g., pets, iRobots, fans, etc.). This recognition includes obtaining a TSCI from a wireless signal transmitted by a Type 1 device and received by a Type 2 device. The action is attributed to an object (e.g., a human, a non-human, a pet, a robot, a vacuum cleaner, a fan, etc.). This recognition also includes calculating an autocorrelation function (ACF) based on the TSCI (e.g., based on a first feature of each CI, such as the magnitude, magnitude squared, a function of the magnitude, phase, component magnitude, component magnitude squared, component phase, or component magnitude function). The TSCI may be sampled at a high rate (the "TSCI rate," e.g., 10,000 Hz), while the ACF may be repeatedly calculated at another rate (the "ACF rate," e.g., 1,000 Hz), with each ACF calculated based on the CIs within a time window (e.g., 0.01 seconds) of the TSCI.

[0365] This recognition also involves identifying ACF features (called "ACF features"). ACF features are peaks / valleys, local maxima (peaks), local minima (valleys), zero crossings, mean crossings, inflection points, restricted features, restricted maxima / minima / zero crossings / mean crossings / inflection points, and / or ACF evaluated at specific time lags (e.g., 1 / 2 / ... / k). Based on the ACF features, several derived analyses / quantities / features can be computed, such as statistical features, ACF-based statistics, motion statistics, speed, physical features, gait, gait-related features, time trend features / statistics (e.g., mean, variance, etc.).

[0366] Some analyses derived from ACF features may be ACF-based statistics (statistical features), including ACF peak mean, ACF valley mean, ACF peak-to-valley distance, and ACF valley-to-peak distance, based on one or more ACFs within a time window. For example, ACFs may be calculated at an ACF rate (e.g., 1000 Hz). Each ACF-based statistic may be calculated using a respective time window of ACFs (e.g., a 0.01-second window consisting of 10 ACFs).

[0367] The ACF peak / valley average may be calculated by computing multiple local maxima (in the case of the ACF peak average) or local minima (in the case of the ACF valley average) of the ACF for each time instance, computing multiple "first averages" (e.g., weighted average, arithmetic mean, geometric mean, harmonic mean, percentile, median (50th percentile), maximum (100th percentile), minimum (0th percentile), mode) of the local maxima / minima at the time instance, and computing a "second average" (e.g., another "first average," or filtering, or low-pass filtering) of the multiple first averages at multiple time instances in the respective time window.

[0368] Alternatively, the ACF peak / valley average can be calculated by calculating the kth (e.g., first) local maximum (in the case of the ACF peak average) or the kth (e.g., first) local minimum (in the case of the ACF valley average) of the ACF for each time instance, and then calculating a "second average" of multiple kth (e.g., first) local maxima / minima at multiple time instances within each time window.

[0369] Alternatively, the ACF peak / valley average can be calculated by calculating the global maximum (in the case of the ACF peak average) or global minimum (in the case of the ACF valley average) of the ACF for each time instance, and then calculating a "second average" of multiple global maxima / minima over multiple time instances within each time window. This may be a special case of (a) where the "first average" is the maximum (in the case of the ACF peak average) or minimum (in the case of the ACF valley average).

[0370] The ACF peak / valley interval distance may be calculated by calculating the number of local maxima (in the case of ACF peak interval distance) or local minima (in the case of ACF valley interval distance) of the ACF for each time instance, calculating multiple intervals (e.g., time intervals) each being the time interval / distance between a pair of adjacent / neighboring local maxima / minima of the ACF for the time instance, calculating a "first average" of the multiple intervals for the time instance, and calculating the ACF peak / valley interval distance as a "second average" (weighted average, or filtered, or low-pass filtered, etc.) of the multiple first averages at the multiple time instances within the respective time window.

[0371] Another analysis derived from the ACF features may be motion statistics (MS), which may be calculated based on the ACF for each time instance (e.g., ACF at a fixed time delay T1). For example, when T1=1, the MS may be the TRRS / inner product / correlation between two temporally adjacent CIs. When T1=k, the MS may effectively be the TRRS based on the TSCI downsampled by a factor of k. MS-based statistics, such as the MS mean and MS variance, may be calculated over a time window.

[0372] The motion statistics average (MS average) may be calculated by calculating a "second average" (such as a first average, filtered, or low-pass filtered) of all motion statistics for time instances within the time window.

[0373] The behavioral statistics variance (MS variance) may be calculated by calculating a "variability measure" of all behavioral statistics for time instances within a time window. The variability measure may include any of the following: variance, standard deviation, variation, derivative, slope, total variation, absolute variation, squared variation, spread, dispersion, variance, deviation, absolute deviation, squared deviation, total deviation, divergence, range, interquartile range, skewness, kurtosis, L-moment, coefficient of variation, interquartile coefficient of dispersion, mean absolute difference, Gini coefficient, relative mean difference, median absolute deviation, mean absolute deviation, distance standard deviation, coefficient of dispersion, entropy, variance-to-mean ratio, maximum-to-minimum ratio, dispersion measure, regularity measure, similarity measure, likelihood, probability distribution function, sampling distribution, and / or moment generating function.

[0374] A specific analysis derived from the ACF features can be the velocity of the object's motion. Each velocity can be calculated based on the local maxima, local minima, zero-crossing points, mean crossing points, inflection points, or the derivative function (i.e., instantaneous slope) of the ACF. This results in a time series of velocity (TSSP). The sampling frequency of the TSSP may be lower than that of the TSCI. Based on the TSSP, several statistics may be further derived / calculated / obtained. Such velocity-based statistics may include several gait statistics (e.g., presence / absence of walking, stride cycle time, step cycle time, N-cycle time, stride cycle distance, step cycle distance, N-cycle distance) and / or several simple velocity statistics (e.g., velocity average, velocity variability measure, velocity percentile, velocity histogram).

[0375] Velocity features ("velocity features") can be calculated, including any of the TSSP peaks (local maxima), valleys (local minima), zero crossings, mean crossings, and / or inflection points. Peaks, valleys, zero crossings, mean crossings, and inflection points are processed by removing "false peaks," "false valleys," "false zero crossings," "false mean crossings," and "false inflection points." False peaks and false valleys may be removed in pairs (i.e., one false peak may be removed along with one adjacent false valley). False zero crossings / mean crossings / inflection points may be removed in pairs (i.e., two immediately adjacent zero crossings / mean crossings / inflection points may be removed together). Mean crossings may be calculated as the zero crossings of the mean-subtracted TSSP.

[0376] One gait statistic is the presence of gait behavior in the object's movement, which can be calculated based on the TSSP velocity features (or simply the TSSP). Gait behavior may be present when velocity features, such as peaks, valleys, zero crossings, mean crossings, and / or inflection points, exhibit rhythmic (e.g., periodic or near-periodic) behavior. For valid presence of gait behavior, the count of peaks, valleys, zero crossings, mean crossings, or inflection points may be upper bounded (by a respective first threshold) and / or lower bounded (by a respective second threshold).

[0377] If walking motion is detected (i.e., walking is present), gait statistics (e.g., stride cycle time, step cycle time, stride length, step length) may be calculated based on the velocity features of the TSSP.

[0378] Gait statistics may be several "intervals" associated with rhythmic movement, with each interval calculated as the time distance or interval between adjacent pairs of peaks, valleys, zero crossings, average crossings, or inflection points. An interval may be associated with a stride (e.g., between two adjacent local maxima or between two adjacent local minima) or a step (e.g., between two adjacent local maxima and minima, between two adjacent inflection points, or between two adjacent average crossings). For example, a first interval is the interval between the first and second local maxima, and a second interval is the interval between the second and third local maxima. An average interval is the "first average" of the intervals.

[0379] Another gait statistic may be several "interval distances" (e.g., stride length, step length) corresponding to several "intervals," where each "interval distance" may be calculated as the distance traversed by an object in the corresponding interval. An interval distance may be calculated as the product of the corresponding interval and a "representative speed" within the interval (e.g., a "first average" or "second average" of the speed). Alternatively, an interval may be subdivided into two or more discontinuous subintervals (e.g., a stride length may be subdivided into two steps). A subinterval distance may be calculated as the product of a subinterval and a "representative speed" within the subinterval. An interval distance may be calculated as the sum or weighted sum of the corresponding subinterval distances. An average interval distance may be the "first average" of several interval distances.

[0380] Another gait statistic is stride cycle time, which is calculated as the "second average" of the interval between pairs of immediately adjacent peaks, or between pairs of immediately adjacent valleys, or between every other pair of zero crossings / mean crossings / inflection points (e.g., between the 1st and 3rd zero crossings, or between the kth and (k+2)th zero crossings).

[0381] The step cycle time may be calculated as half the stride cycle time, or based on one-quarter of the "second average" of the interval between adjacent zero crossing / average crossing / inflection point pairs (e.g., between the first and second zero crossings, or between the kth and (k+1)th zero crossings), or the interval between a local maximum and an adjacent local minimum.

[0382] The time for N cycles can be calculated as the interval between N peaks, or N valleys, or 2N zero crossings / mean crossings / inflection points.

[0383] The step cycle distance, stride cycle distance, and / or N-cycle distance may be calculated based on the interval distance. The step cycle distance, stride cycle distance, or N-cycle distance may be defined as the distance traversed by an object motion in a step cycle time, stride cycle time, or N-cycle time. The step cycle distance, stride cycle distance, or N-cycle distance may be calculated based on the sum of the interval distances associated with all intervals within the corresponding step cycle time, stride cycle time, or N-cycle time. The step cycle distance, stride cycle distance, or N-cycle distance may be averaged over a time period using a first average or a second average.

[0384] Simple speed statistics (e.g., second average speed, speed "variability index," 25th percentile speed, 50th percentile speed / median speed, 75th percentile speed) may be calculated over a period of time. A histogram of speeds within the period may also be calculated. Based on the histogram, any speed percentile or variability index may be calculated.

[0385] The monitoring tasks can be performed based on ACF-based statistics, MS-based statistics, and speed-based statistics (including both gait statistics and simple speed statistics). The monitoring tasks can include at least one of classification / detection / recognition / identification / presence / absence / appearance / disappearance / ADL (Activities of Daily Living) of at least two of the following: human, human adult, elderly, human child, human baby, non-human, non-human pet, non-human robot, and non-human robot. Tasks can include gesture recognition, tracking, orientation, localization, and navigation.

[0386] The task is to detect the motion of an (unknown) object and distinguish / recognize / detect / classify / identify the object producing the detected motion from a number of possible / candidate objects (e.g. there could be four candidate / possible objects: a human, a pet, a robot, and a fan).

[0387] Some subtasks of the task are to perform pairwise differentiation / recognition / classification / detection between all object pair combinations. For example, in this example with four candidate objects, pairwise detection can be performed for all pair combinations including (a) human vs. pet, (b) human vs. robot, (c) human vs. fan, (d) pet vs. robot, (e) pet vs. fan, and (f) robot vs. fan.

[0388] For each pairwise detection, a pairwise detection score may be calculated. For example, the pairwise detection score for object 1 versus object 2 tends to be positive (or above a threshold) if the object is object 1, and tends to be negative (or below a threshold) if the object is object 2. Then, the object is determined / detected / classified / identified as object 1 if the score is above the threshold, and determined / detected / classified / identified as object 2 otherwise.

[0389] For each potential / candidate object (e.g., human), a respective per-object detection score for the potential / candidate object (e.g., human) may be calculated based on all pairwise detection scores associated with that object (i.e., with the candidate object being either object 1 or object 2 in the object 1 vs. object 2 pairwise detection). The per-object score may be a weighted aggregation (e.g., weighted sum, weighted product) of the pairwise scores. In the weighted aggregation, the score weight for any pair may be positive if the potential object (human) is object 1 in the object 1 vs. object 2 pairwise detection and negative if it is object 2.

[0390] In this example, a per-object detection score for a human can be calculated based on the pairwise detection scores for (a) human vs. pet, (b) human vs. robot, and (c) human vs. fan. The per-object detection score can be a weighted aggregation of the pairwise detection scores, with (a), (b), and (c) being positively weighted.

[0391] A per-object detection score for the pet can be calculated based on three pairwise scores associated with the pet: (a) human vs. pet, (d) pet vs. robot, and (e) pet vs. fan. The per-object detection score can be a weighted sum of the pairwise detection scores, with positive weights for (d) and (e) and a negative weight for (a).

[0392] The robot's per-object detection score may be calculated based on three pairwise scores associated with the robot: (b) human vs. robot, (d) pet vs. robot, and (f) robot vs. fan. The per-object detection score may be a weighted sum of the pairwise detection scores, with a positive weight on (f) and negative weights on (b) and (d).

[0393] A per-object detection score for the fan may be calculated based on three pairwise scores associated with the robot: (c) human vs. fan, (e) pet vs. fan, and (f) robot vs. fan. The per-object detection score may be a weighted sum of the pairwise detection scores, with negative weights on (c), (e), and (f).

[0394] Each of the per-object detection scores may be further weighted (e.g., based on some prior information): for example, if it is known that there is a high probability of it being a human, then a separate positive weight may be applied to that per-object detection score, making it larger.

[0395] The object associated with the detected motion may be determined to be the candidate object for which the per-object detection score is the greatest among all the candidate objects.

[0396] To calculate a per-object detection score y for all candidate objects (e.g., y may be a 4x1 vector in this example with four candidate objects), the pairwise scores may be organized to form a vector x (e.g., x may be a 6x1 vector in this example with six pairwise detection scores). The corresponding weights for calculating each weighted sum may be organized into a matrix A (e.g., A may be a 4x6 matrix) such that y is calculated as the product of A and x (i.e., y = Ax).

[0397] The monitoring task (or each subtask) may be performed based on machine learning, classifiers, support vector machines (SVMs), supervised learning, unsupervised learning, semi-supervised learning, clustering, feature extraction, feature learning, principal component analysis, eigendecomposition, frequency decomposition, time decomposition, time-frequency decomposition, training, discriminative training, supervised training, unsupervised training, semi-supervised training, neural networks, and / or deep learning, and includes multiple inputs: (a) ACF peak mean, ACF valley mean, ACF peak interval distance, ACF valley interval distance. (b) ACF-based statistics (or ACF statistics) such as (a) MS-based statistics (or movement statistics) such as MS mean and MS variance, (c) speed-based statistics (or speed statistics) such as speed mean, speed variance, speed percentile, speed 50th percentile, speed median, speed 0th percentile, minimum speed, speed 100th percentile, maximum speed, speed 25th percentile, and speed 75th percentile, and (d) gait statistics such as gait presence, interval, stride cycle time, step cycle time, N-cycle time, interval distance, stride cycle length, step cycle length, and N-cycle distance. There are four types of statistics, including (a) ACF-based statistics, (b) MS-based statistics, (c) speed-based statistics, and (d) gait statistics.

[0398] A Type 1 or Type 2 device may have multiple antennas so that multiple links exist. Multiple TSCIs are obtained based on the received wireless signals, with each TSCI obtained from a respective link. Based on each TSCI, a respective ACF may be calculated, and (a) ACF-based statistics, (b) MS-based statistics, (c) speed-based statistics, and (d) walking statistics may be calculated. If one of the links is faulty / failed / noisy (e.g., its RX antenna is faulty), some or all of the ACF-based statistics, MS-based statistics, speed-based statistics (including the speed itself), and walking statistics of the faulty link may be replaced / augmented / supplemented / mixed with the respective statistics from other good / non-failed / non-noisy link(s).

[0399] The classifier may use N inputs, including (a) ACF-based statistics, (b) MS-based statistics, (c) speed-based statistics, and (d) gait statistics. An alternative classifier with Nk inputs may be trained such that if corresponding k inputs out of the N inputs are missing or bad, the alternative classifier is used instead.

[0400] From the TSCI obtained from the received radio (sounding) signal communicated between the Type 1 device (TX) and the Type 2 device (RX), the autocorrelation function (ACF) of the CI of the TSCI may be calculated based on several features of the CI of the TSCI (referred to as CI features, e.g., magnitude, magnitude squared, magnitude function, and phase). At least one feature of the ACF or a function of the ACF (e.g., a differential function of the ACF) (referred to as ACF feature, e.g., local maximum, local minimum, zero crossing, mean crossing, and inflection point) may be calculated / obtained. At least four statistics (from four different types) may be calculated, each based on a respective ACF feature of the ACF or ACF function. Some or all of the ACF features may be the same. There may be at least one statistic from each of four types of statistics: a) ACF statistics (or ACF-based statistics), (b) motion statistics (or MS-based statistics), (c) speed statistics (or speed-based statistics), and (d) gait statistics. If motion is detected, the object may be classified into one of a number of candidate object classes (e.g., human, pet (non-human), robot (non-human), fan (non-human)). In some embodiments, motion statistics (type (b) statistics) may be used to detect motion. Motion may be detected when a motion statistic value, or an accumulation thereof, exceeds a threshold.

[0401] In some embodiments, an ACF may be calculated for each sampling moment of a CI of a TSCI, with each ACF calculated based on the CIs within a corresponding sliding time window. Each ACF may be calculated based on CI features (e.g., magnitude, magnitude squared, phase) of each CI within a corresponding sliding time window. In some embodiments, at least one ACF feature (e.g., local maximum, local minimum, zero crossing, mean crossing, inflection point) of each ACF or a function of each ACF may be calculated. In some embodiments, ACF statistics (type (a) statistics) and behavior statistics (type (b) statistics) may be calculated based on several first ACF features and second ACF features of the ACF or a function of the ACF in the corresponding sliding time window, respectively.

[0402] In some embodiments, a time series of the "speed" of an object's motion may be calculated based on some ACF feature (e.g., a local maximum / minimum of the ACF, a local maximum / minimum of the derivative (or slope) of the ACF, or a function of the ACF). Speed ​​statistics (type (c) statistics) may be calculated based on the speed within a time window of the TSS. The TSS may be analyzed to detect the presence of walking (to search for periodic behavior that matches walking). If walking is detected, walking statistics (type (d) statistics) may be calculated based on the TSS. Several additional candidate object classes, including "pet," "robot," and "fan."

[0403] In some embodiments, there may be multiple (e.g., N) candidate object classes. Many (to be precise, N(N-1) / 2) pairings of two candidate object classes may be discovered / established. Many (to be precise, N(N-1) / 2) pairwise detections may be performed, one for each pair / pairing. Many pairwise detection scores may be calculated, one for each pairwise detection / pair / pairing. For each pairwise detection between a first object and a second object, the pairwise detection score tends to be positive (or at least greater than a threshold) if the object is the first object, and tends to be negative (or at least less than a threshold) if the object is the second object. For each particular candidate object class, a detection score for each object may be calculated based on all pairwise detection scores associated with the particular candidate object class (i.e., with the particular candidate object class being either the first object or the second object in the associated pairwise detection).

[0404] In some embodiments, the per-object detection score for a particular candidate object class may be a weighted aggregation (e.g., weighted sum, weighted product) of all pairwise detection scores. For any pairwise detection score where the particular candidate object class does not participate in the associated pairwise detection, the weight may be zero. For any pairwise detection score where the particular candidate object class is the first object class in the associated pairwise detection, the weight may be positive. For any pairwise detection score where the particular candidate object class is the second object class in the associated pairwise detection, the weight may be negative. In some embodiments, when motion is detected, the object is classified into the candidate object class with the largest per-object detection score.

[0405] In some embodiments, tasks may be performed based on the detected motion and object classification. In a security task, if a venue is expected to be empty, an intruder may be detected if motion is detected and the object is classified as human. If the object is not classified as human, the detected motion may be considered a robot vacuum, pet, or fan and ignored.

[0406] In some embodiments, the TSCI may be sampled at a high rate. The TSCI is resampled to a lower rate, resulting in a time series of resampled CIs (TSRCI). Motion detection may be less demanding, but object classification may be more demanding. Thus, object classification is performed using the high-rate TSCI, and motion detection is performed using the low-rate TSRCI.

[0407] In some embodiments, the sampling rate of the TSCI may be greater than that of the TSRCI. The sampling rate of the TSCI may be five times or more than that of the TSRCI. Several possible CIs are: CSI, CIR, CFR, etc. The TSCIs may be pre-processed. Each ACF is calculated based on the features of the CI (e.g., magnitude, magnitude squared, phase).

[0408] In some embodiments, motion may be detected when the magnitude of the MS, or the cumulative magnitude of the MS, is greater than a threshold.

[0409] The following numbered sections provide examples relating to wireless human-non-human motion detection and recognition.

[0410] Section A1. A method / apparatus / system / software for wireless human-non-human motion detection, comprising: a transmitter configured to transmit a wireless signal affected by an object's motion through a wireless multipath channel of a venue; a receiver configured to receive the wireless signal through the wireless multipath channel of the venue, the received wireless signal being different from the transmitted wireless signal due to the wireless multipath channel and the object's motion; and a processor configured to: obtain time series channel information (TSCI) of the wireless multipath channel based on the received wireless signal; and calculate an autocorrelation function (ACF) based on the TSCI. calculate at least one ACF feature of the ACF or a function of the ACF, calculate ACF statistics based on a first ACF feature of the ACF or the function of the ACF, calculate motion statistics associated with the motion of the object based on a second ACF feature of the ACF or the function of the ACF, calculate speed statistics associated with the speed of the motion of the object based on a third ACF feature of the ACF or the function of the ACF, and calculate gait statistics associated with the walking motion of the object based on a fourth ACF feature of the ACF or the function of the ACF, and if motion is detected, classify the object associated with the detected motion as either human or non-human based on the ACF statistics, the motion statistics, the speed statistics, and the gait statistics.

[0411] Paragraph A2. The method / apparatus / system / software for wireless human-non-human motion detection of Paragraph A1, including a processor further configured to detect object motion based on motion statistics and classify the object associated with the detected motion as one of human or non-human based on ACF statistics, motion statistics, speed statistics, and gait statistics.

[0412] Paragraph A3. The method / apparatus / system / software for wireless human-non-human activity detection of paragraphs A1-A2, including a processor further configured to: calculate a time series ACF (TSACF) for each ACF based on the CI within the first sliding time window of each of the TSCIs; and calculate at least one ACF feature for each ACF or a function of each ACF.

[0413] Paragraph A4. The method / apparatus / system / software for wireless human-non-human motion detection of Paragraph A3, including a processor further configured to calculate ACF statistics based on the first ACF feature of each ACF or function of the ACF in the second time window of the TSACF.

[0414] Paragraph A5. The method / apparatus / system / software for wireless human-non-human motion detection of paragraphs A3-A4, including a processor further configured to: calculate a time series of velocities (TSS) of object motion, each respective velocities based on a third ACF feature of each ACF or function of ACF in a third sliding time window of each of the TSACFs; and calculate velocity statistics based on each velocities in a fourth time window of the TSS.

[0415] Paragraph A6. The method / apparatus / system / software for wireless human-non-human motion detection of Paragraph A5, including a processor further configured to detect the presence of walking motion of an object based on the TSS.

[0416] Paragraph A7. The method / apparatus / system / software for wireless human-non-human motion detection of paragraphs A5-A6, including a processor further configured to: calculate gait statistics based on each speed in a fifth time window of the TSS, wherein the fourth ACF feature is the third ACF feature.

[0417] Clause A8. The method / apparatus / system / software for wireless human-non-human motion detection of clauses A2-A7, further comprising a processor configured to, when motion is detected, classify an object associated with the detected motion as one of a human, an adult, a child, a baby, a non-human, a pet, a robot, a machine, a fan, multiple humans, or multiple non-humans based on ACF statistics, motion statistics, speed statistics, and gait statistics.

[0418] Clause A9. The method / apparatus / system / software for wireless human-non-human motion detection of clause A8, including a processor further configured to: determine a plurality of candidate object classes for the object; determine a plurality of pairings of candidate object classes from the plurality of candidate object classes; perform a plurality of pairwise detections based on ACF statistics, motion statistics, speed statistics, and gait statistics, each pairwise detection being for a respective pairing of candidate object classes; and, when motion is detected, classify the object into one of the plurality of candidate object classes based on the plurality of pairwise detections.

[0419] Clause A10. The method / apparatus / system / software for wireless human-non-human motion detection of clause A9, further including a processor configured to calculate a plurality of pairwise detection scores based on ACF statistics, motion statistics, speed statistics, and gait statistics, each pairwise detection score being for a respective pairwise detection between a respective first candidate object class and a respective second candidate object class, each score being generally greater when the object is a first candidate score than when the object is a second candidate score, and when motion is detected, classify the object into one of the plurality of candidate object classes based on the plurality of pairwise detection scores.

[0420] Clause A11. The method / apparatus / system / software for wireless human-non-human action detection of clause A10, further comprising: a processor configured to calculate a plurality of per-object detection scores, wherein each per-object detection score for each particular candidate object class is calculated as a weighted aggregation of all of the weighted plurality of pairwise detection scores with (a) a weight of zero if each particular candidate object class is neither the first object nor the second object in the associated pairwise detection, (b) a positive weight if each particular candidate object class is the first candidate object class in the associated pairwise detection, or (c) a negative weight if each particular candidate object class is the second candidate object class in the associated pairwise detection, and if action is detected, classify the object into one of the plurality of candidate object classes based on the plurality of per-object detection scores.

[0421] Clause A12. The method / apparatus / system / software for wireless human-non-human motion detection of clause A11, including a processor further configured to classify the objects into the candidate object class having the highest per-object detection score when motion is detected.

[0422] Paragraph A13. The method / apparatus / system / software for wireless human-non-human motion detection of paragraphs A2-A12, including a processor further configured to perform a task based on the detected motion and a classification of an object associated with the detected motion.

[0423] Clause A14. The method / apparatus / system / software for wireless human-non-human motion detection of any preceding clause, including a processor further configured to detect an intruder if motion is detected and the object is classified as human.

[0424] Clause A15. The method / apparatus / system / software for wireless human-non-human motion detection of any preceding clause, including a processor further configured to: resample the TSCI to obtain a time series of resampled CIs (TSRCI); and calculate motion statistics based on the TSRCI.

[0425] Clause A16. The method / apparatus / system / software for wireless human-non-human motion detection of clause A15, including a processor further configured such that the sampling rate of TSRI is greater than the sampling rate of TSRCI.

[0426] Paragraph A17. The method / apparatus / system / software for wireless human-non-human activity detection of paragraph A16, including a processor further configured such that the sampling rate of the TSCI is at least five times greater than the sampling rate of the TSRCI.

[0427] Clause A18. The method / apparatus / system / software for wireless human-non-human activity detection of any preceding clause, wherein any channel information (CI) includes at least one of channel state information (CSI), channel impulse response (CIR), channel frequency response (CFR), signal power, magnitude, phase, a function of magnitude, a function of phase, magnitude squared, average magnitude, weighted magnitude, average phase, weighted phase, phase difference, or any of the above features.

[0428] Section A19. The method / device / system / software for wireless human-non-human motion detection of any of the preceding sections, including a processor further configured to process the TSCI.

[0429] Clause A20. The method / device / system / software for wireless human-non-human motion detection of any of the preceding clauses, including the processor configured to: calculate an ACF based on CI features for each CI of the TSCI, where the CI features for a CI include: CI magnitude, CI phase, CI component magnitude, CI component phase, CI magnitude function, CI magnitude squared, CI phase function, CI component magnitude function, CI component magnitude squared, or CI component phase function.

[0430] Clause A21. The method / apparatus / system / software for wireless human-non-human motion detection of clause A20, wherein the component includes one of a tap of a CIR, a subcarrier of a CFR, or a component of a CSI.

[0431] Clause A22. The method / apparatus / system / software for wireless human-non-human motion detection of any of the preceding clauses, wherein the motion statistics include at least one of the following pairwise quantities between the two channel information (CI): difference, distance, weighted distance, Euclidean distance, absolute distance, angular distance, graph distance, statistical distance, distance metric, L_1 norm, L_2 norm, L_k norm, distance score, similarity, similarity score, dot product, dot product, correlation, correlation coefficient, correlation index, autocorrelation, autocovariance, cross-correlation, cross-covariance, or an aggregation thereof.

[0432] Clause A23. The method / apparatus / system / software for wireless human-non-human motion detection of any preceding clause, wherein the ACF statistics include at least one of the following: ACF peak average, ACF valley average, ACF zero-crossing average, ACF mean crossing average, ACF inflection point average, ACF(t) average, ACF peak variance, ACF valley variance, ACF zero-crossing variance, ACF mean crossing variance, ACF inflection point variance, ACF(t) variance, ACF peak-to-peak interval, ACF valley-to-valley interval, ACF zero-crossing interval, ACF mean crossing interval, ACF inflection point interval, ACF peak interval distance, ACF valley interval distance, ACF peak-to-valley interval distance, ACF zero-crossing interval distance, ACF mean crossing interval distance, or ACF inflection point interval distance, where ACF(t) is the ACF or a function of the ACF evaluated at a particular argument value t.

[0433] Clause A24. The method / apparatus / system / software for wireless human-non-human motion detection of any of the preceding clauses, wherein the speed statistics include at least one of the following speeds: histogram, mean, percentile, 50th percentile, median, percentile, 0th percentile, minimum, 100th percentile, maximum, 25th percentile, 75th percentile, trimmed mean, conditional mean, weighted mean, weighted median, mode, sum, weighted sum, product, weighted product, arithmetic mean, geometric mean, harmonic mean, variance, standard deviation, variation, variance, deviation, derivative, slope, spread, dispersion, range, skewness, kurtosis, L-moment, entropy, variance-to-mean ratio, maximum-to-minimum ratio, regularity, similarity, likelihood, correlation, covariance, autocorrelation, autocovariance, or a sliding scale of any of the above.

[0434] Clause A25. The method / apparatus / system / software for wireless human-non-human motion detection of any of the preceding clauses, wherein the gait statistics include at least one of the following: presence of walking, interval, stride cycle time, step cycle time, N cycle time, interval distance, stride cycle distance, stride length, step cycle distance, step length, N cycle distance, asymmetry between even and odd cycles, asymmetry between modulo 1 and modulo 2 cycles, asymmetry between modulo 1 and modulo 3 cycles, asymmetry between modulo 1 and modulo 4 cycles, asymmetry between modulo 2 and modulo 3 cycles, asymmetry between modulo 2 and modulo 4 cycles, or asymmetry between modulo 3 and modulo 4 cycles.

[0435] Clause A26. The method / apparatus / system / software for wireless human-non-human motion detection of any of the preceding clauses (particularly clause A2) including a processor further configured to detect motion of the object when a magnitude of the motion statistic, or an accumulation of the magnitudes of the motion statistics, exceeds a threshold.

[0436] Clause A27. The method / apparatus / system / software for wireless human-non-human motion detection of Clause A26, wherein the accumulation of motion statistics includes at least one of the following: sum, integral, weighted sum, product, weighted product, average, weighted mean, mean value, arithmetic mean, geometric mean, harmonic mean, conditional mean, trimmed mean, statistic, median, weighted median, percentile, mode, maximum, minimum, variance, deviation, variation, sliding average, sliding mean, sliding weighted mean, sliding median, sliding variance, sliding deviation, or sliding statistic.

[0437] Clause A28. The method / apparatus / system / software for wireless human-non-human motion detection of any of the preceding clauses, wherein at least one ACF feature comprises any of the following: a global maximum, a global minimum, a local maximum, a local minimum, a zero crossing, a mean crossing, an inflection point, a kth local maximum, a kth minimum, a kth zero crossing, a kth mean crossing, a kth inflection point, an ACF or a function of an ACF related to any of the above, a set of adjacent local maxima, a set of adjacent local minima, a set of adjacent local maxima and local minima, a set of adjacent zero crossings, a set of adjacent mean crossings, a set of adjacent inflection points, a histogram of any of the above, a statistic related to a histogram, a mean, a conditional mean, a trimmed mean, a median, a mode, a percentile, a variance, a deviation, a variability, a weighted mean, an arithmetic mean, a time argument of ... mean, geometric mean, harmonic mean, presence of rhythm over time, periodic behavior, repetitive behavior, frequency, period, a function of one of the above pairs, the difference of one of the above pairs, the time interval between one of the above pairs, or an ACF or a function of an ACF evaluated at a specific time argument, where the ACF function includes any of the following: linear function, piecewise linear function, nonlinear function, polynomial function, exponential function, logarithmic function, trigonometric function, transcendental function, absolute function, magnitude function, index function, threshold function, quantization function, derivative function, first derivative, second derivative, third derivative, integral function, single integral, double integral, triple integral, or at least one of the functions obtained by the following filtering of the ACF: low-pass filtering, band-pass filtering, high-pass filtering, smoothing, weighted averaging.

[0438] Clause A29. The method / apparatus / system / software for wireless human-non-human motion detection of any of the preceding clauses, further comprising a processor configured to, when a motion is detected, classify an object associated with the detected motion based on ACF statistics, motion statistics, speed statistics, and gait statistics using at least one of the following: support vector machine (SVM), SVM with a linear kernel function, SVM with a nonlinear kernel, SVM with a polynomial kernel, SVM with a Gaussian kernel, SVM with a sigmoid kernel, neural network (NN), deep neural network (DNN), recurrent neural network (RNN), perceptron, multilayer perceptron (MLP), feedforward neural network, recurrent Hopfield network, radial basis function neural network, neural backpropagation, convolutional neural network (CNN), AlexNet, LeNet, residual neural network (ResNet), recurrent neural network, generative adversarial network (GAN), cellular neural network, Dan Ciresan Net, Overfeat, VGG, LeNet5, Network-in-Network, GoogLeNet, FaceNet, Inception, Bottleneck Layer, SqueezeNet, ENet, Fully Connected Neural Networks, Deconvolutional Neural Networks, Long Short-Term Memory (LSTM), BiLSTM, Modular Neural Networks, Sequence-to-Sequence Models, Siamese Architectures, Gated Recurrent Units, Boltzmann Machines, Deep Belief Networks, Transformers, ViT, Gated Recurrent Units (GRU), CNN+GRU, Hybrids, Combining Two or More Neural Networks, Supervised Learning, Unsupervised Learning, Semi-Supervised Learning, Deep Learning, Random Search, Bayesian Optimization, Evolutionary Methods, Reinforcement Learning, Gradient-Based Methods, Centralized Network Management, Decentralized Network Management, or Hierarchical Network Management.

[0439] Clause A30. The method / device / system / software for wireless human-non-human activity detection of clause A13, wherein the task includes at least one of the following: (a) object, multiple objects, vehicle, machine, tool, human, baby, elderly, patient, pet, presence, proximity, activity, daily activity, well-being, vital signs, breathing, heart rate, health status, sleep, sleep stage, walking, gait, gesture, movement, fall, danger, intrusion, security, safety, life threat, emotion, movement, motion, degree of movement, periodic movement, repetitive movement, stationary movement, periodic stationary movement, regular movement, transient movement, sudden movement, suspicious. (b) Sensing tasks, including any of monitoring, sensing, detection, recognition, estimation, verification, identification, authentication, classification, localization, guidance, navigation, tracking, or counting of at least one of: motion, irregularities, trends, human biometrics, environmental informatics, rooms, venues, indoor venues, or cars in a parking lot; (c) analysis, including machine learning, supervised learning, unsupervised learning, semi-supervised learning, deep learning, clustering, feature extraction, feature learning, principal component analysis, eigendecomposition, frequency decomposition, time decomposition, time-frequency decomposition, function decomposition, and other decompositions. (c) IoT tasks, including any of the following: any sensing task, energy management, power transfer, wireless power transfer, device or system activation, alarm, siren, speaker, television, entertainment system, camera, heater, coffee maker, cooking device, cleaning device, household device, security system, access system, HVAC system, ventilation system, lighting system, gaming system; or (d) transmission, processing, storage, or analysis of data, parameters, analysis, or derived data, management, configuration, upgrade, adjustment, broadcast, synchronization, networking, communication, encryption, protection, compression, storage, database, archiving, query, cloud computing, presentation.Any other task involving augmented reality or virtual reality.

[0440] The following numbered sections provide examples relating to wireless human-non-human motion detection and recognition.

[0441] Section 3813_1. A system for wireless human-non-human motion detection, the system including: a transmitter configured to transmit a wireless signal through a wireless multipath channel of a venue; a receiver configured to receive the wireless signal through the wireless multipath channel of the venue, wherein the received wireless signal differs from the transmitted wireless signal due to the wireless multipath channel being affected by the motion of an object within the venue; and a processor configured to: obtain time series channel information (TSCI) of the wireless multipath channel based on the received wireless signal; calculate an autocorrelation function (ACF) based on the TSCI; and calculate the ACF or ACF. calculate at least one ACF feature of a function of the ACF, calculate ACF statistics based on the ACF or a first ACF feature of the function of the ACF, calculate motion statistics associated with a motion of the object based on the ACF or a second ACF feature of the function of the ACF, calculate speed statistics associated with a speed of the motion of the object based on a third ACF feature of the ACF or a function of the ACF, calculate gait statistics associated with a walking motion of the object based on a fourth ACF feature of the ACF or a function of the ACF, and in response to determining that a motion has been detected, classify the object associated with the detected motion as either human or non-human based on the ACF statistics, the motion statistics, the speed statistics, and the gait statistics.

[0442] Item 3813_2. The system of Item 3813_1, wherein the processor is further configured to detect motion of the object based on the motion statistics.

[0443] Term 3813_3. The system of Term 3813_1, wherein the processor is further configured to calculate a time series of ACFs (TSACFs), each ACF calculated based on a CI within a first sliding time window of each of the TSACFs, and to calculate at least one ACF feature for each ACF or a function of each ACF.

[0444] Term 3813_4. The system of Term 3813_3, wherein the processor is further configured to calculate ACF statistics based on the first ACF characteristic of each ACF or function of each ACF within a second sliding time window of the TSACF.

[0445] Clause 3813_5. The system of clause 3813_4, further configured such that the processor calculates a time series of velocities (TSS) of the object's motion, each velocity calculated based on a third ACF feature of a respective ACF or function of the ACF in a third sliding time window of each of the TSACFs, and calculates velocity statistics based on the respective velocities in a fourth sliding time window of the TSS.

[0446] Item 3813_6. The system of Item 3813_5, wherein the processor is further configured to detect the presence of a walking motion of the object based on the TSS.

[0447] Term 3813_7. The system of Term 3813_6, wherein the processor is further configured to calculate gait statistics based on each speed within a fifth sliding time window of the TSS, and wherein the fourth ACF feature is the same as the third ACF feature.

[0448] Clause 3813_8. The system of Clause 3813_7, wherein the processor is further configured, in response to determining that a motion is detected, to classify an object associated with the detected motion as one of a human, an adult, a child, a baby, a non-human, a pet, a robot, a machine, a fan, multiple humans, or multiple non-humans based on the ACF statistics, the motion statistics, the speed statistics, and the gait statistics.

[0449] Clause 3813_9. The system of Clause 3813_8, wherein the processor is further configured to determine a plurality of candidate object classes for the object, determine a plurality of pairings of candidate object classes from the plurality of candidate object classes, and perform a plurality of pairwise detections based on the ACF statistics, the motion statistics, the speed statistics, and the gait statistics, each pairwise detection being performed for a respective pairing of candidate object classes, and in response to determining that motion is detected, classify the object into one of the plurality of candidate object classes based on the plurality of pairwise detections.

[0450] Clause 3813_10. The system of Clause 3813_9, wherein the processor is further configured to calculate a plurality of pairwise detection scores based on the ACF statistics, the motion statistics, the speed statistics, and the gait statistics, each pairwise detection score being calculated for a respective pairwise detection between a respective first candidate object class and a respective second candidate object class, and in response to determining that motion is detected, classify the object into one of the plurality of candidate object classes based on the plurality of pairwise detection scores.

[0451] Clause 3813_11. The system of Clause 3813_10, wherein the processor is further configured to calculate a plurality of per-object detection scores, each per-object detection score calculated for a respective particular candidate object class, each per-object detection score calculated as a respective weighted aggregation of all the plurality of pairwise detection scores, each pairwise detection score of t...

Claims

1. 1. A system for wireless human-non-human motion detection, comprising: a transmitter configured to transmit a wireless signal over a wireless multipath channel of the venue; a receiver configured to receive the wireless signal over the wireless multipath channel of the venue, the received wireless signal differing from the transmitted wireless signal due to the wireless multipath channel being affected by the motion of objects within the venue; 1. A processor comprising: Obtaining a time series of channel information (TSCI) of the wireless multipath channel based on the received wireless signal; Calculating an autocorrelation function (ACF) based on the TSCI; Calculating at least one ACF feature, which is the ACF or a function of the ACF; calculating an ACF statistic based on a first ACF feature of the ACF or the function of the ACF; calculating motion statistics associated with the motion of the object based on a second ACF feature of the ACF or the function of the ACF; calculating a velocity statistic related to a velocity of the movement of the object based on a third ACF feature of the ACF or the function of the ACF; calculating gait statistics associated with a walking motion of the object based on a fourth ACF feature of the ACF or the function of the ACF; In response to determining that the motion is detected, classifying the object associated with the detected motion as either human or non-human based on the ACF statistics, the motion statistics, the speed statistics, and the gait statistics. and the processor configured to Wireless human-non-human motion detection system.

2. The processor is further configured to detect the motion of the object based on the motion statistics. The system of claim 1 .

3. The processor, calculating a time series of ACFs (TSACFs), each ACF being calculated based on a CI for a first sliding time window for each of the TSCIs; Calculating at least one of the ACF features for each ACF or a function of each ACF; calculating the ACF statistics based on the first ACF characteristics of each ACF or the function of each ACF in a second sliding time window of the TSACF; It is further configured as follows: The system of claim 1 .

4. The processor, calculating a time series of velocities (TSS) of the motion of the object, each of the velocities being calculated based on the third ACF feature of each ACF or the function of each ACF in a third sliding time window of each of the TSACFs; calculating the speed statistics based on each speed in a fourth sliding time window of the TSS; Based on the TSS, the presence or absence of the walking motion of the object is detected. It is further configured as follows: The system of claim 3.

5. The processor, Calculate the gait statistics based on each speed in a fifth sliding time window of the TSS, and the fourth ACF feature is the same as the third ACF feature; In response to determining that the motion has been detected, classify the object associated with the detected motion as one of the human, an adult, a child, a baby, the non-human, a pet, a robot, a machine, a fan, humans, or non-humans based on the ACF statistics, the motion statistics, the speed statistics, and the walking statistics. It is further configured as follows: The system of claim 4.

6. The processor, determining a plurality of candidate object classes for the object; determining a plurality of pairings of candidate object classes from the plurality of candidate object classes; performing a plurality of pairwise detections based on the ACF statistics, the motion statistics, the speed statistics, and the gait statistics, each pairwise detection being performed for a respective pairing of the candidate object classes; In response to determining that the motion has been detected, classifying the object into one of the plurality of candidate object classes based on the plurality of pairwise detections. It is further configured as follows: The system of claim 5.

7. The processor, calculating a plurality of pairwise detection scores based on the ACF statistics, the motion statistics, the speed statistics, and the gait statistics, each pairwise detection score being calculated for a respective pairwise detection between a respective first candidate object class and a respective second candidate object class; In response to determining that the motion has been detected, classifying the object into one of the plurality of candidate object classes based on the plurality of pairwise detection scores. It is further configured as follows: The system of claim 6.

8. The processor, Calculating a detection score for each of a plurality of objects, Each of the per-object detection scores is calculated for each particular candidate object class, Each of the detection scores for each object is calculated as a weighted aggregation of all of the plurality of pairwise detection scores; each respective pairwise detection score of the plurality of pairwise detection scores is weighted (a) with a positive weight if the respective particular candidate object class is a respective first candidate object class in the respective pairwise detection associated with the respective pairwise detection score, (b) with a negative weight if the respective particular candidate object class is a respective second candidate object class in the respective pairwise detection, or (c) with a weight of zero if the respective particular candidate object class is neither the respective first candidate object class nor the respective second candidate object class in the respective pairwise detection; In response to determining that the motion has been detected, classifying the object into one of the plurality of candidate object classes based on a detection score for each of the plurality of objects. It is further configured as follows: The system of claim 7.

9. The processor, responsive to determining that the motion is detected, classifying the object into a candidate object class having a maximum per-object detection score; performing a task based on the detected motion and the classification of the object associated with the detected motion; Detecting an intruder if the motion is detected and the object is classified as a human; re-sampling the TSCI to obtain a time series of resampled CI (TSRCI); Calculate the performance statistics based on the TSRCI It is further configured as follows: The system of claim 8.

10. the sampling rate of the TSCI is at least five times greater than the sampling rate of the TSRCI; The channel information (CI) of any of the TSCI includes at least one of the following: channel state information (CSI), channel impulse response (CIR), channel frequency response (CFR), signal power, magnitude, phase, a function of the magnitude, a function of the phase, the magnitude squared, the average magnitude, the weighted magnitude, the average phase, the weighted phase, the phase difference, or each of the respective features. The system of claim 9.

11. The processor, calculating the ACF based on CI features of each CI of the TSCI, the CI features including a magnitude of the CI, a phase of the CI, a magnitude of a component of the CI, a phase of a component of the CI, a function of the magnitude of the CI, a square of the magnitude of the CI, a function of the phase of the CI, a function of the magnitude of the component of the CI, a square of the magnitude of the component of the CI, or a function of the phase of the component of the CI; The components include one of a tap of a CIR, a subcarrier of a CFR, or a component of a CSI. It is further configured as follows: The system of claim 10.

12. The operational statistics include at least one of the following pairwise quantities between two Channel Information (CI): difference, distance, weighted distance, Euclidean distance, absolute distance, angular distance, graph distance, statistical distance, distance metric, L_1 norm, L_2 norm, L_k norm, distance score, similarity, similarity score, inner product, dot product, correlation, correlation coefficient, correlation index, autocorrelation, autocovariance, cross-correlation, cross-covariance, or a summary of each. The system of claim 11.

13. The ACF statistics include at least one of ACF peak average, ACF valley average, ACF zero cross average, ACF mean cross average, ACF inflection point average, ACF(t) mean, ACF peak variance, ACF valley variance, ACF zero cross variance, ACF mean cross variance, ACF inflection point variance, ACF(t) variance, ACF peak-to-peak interval, ACF valley-to-valley interval, ACF zero cross interval, ACF mean cross interval, ACF inflection point interval, ACF peak interval distance, ACF valley interval distance, ACF peak-to-valley interval distance, ACF zero cross interval distance, ACF mean cross interval distance, or ACF inflection point interval distance, where ACF(t) is the ACF or the function of the ACF evaluated at a particular argument value t. The system of claim 12.

14. The rate statistics include at least one of the following functions of the rates: histogram, mean, percentile, 50th percentile, median, 0th percentile, minimum, 100th percentile, maximum, 25th percentile, 75th percentile, trimmed mean, conditional mean, weighted average, weighted median, mode, sum, weighted sum, product, weighted product, arithmetic mean, geometric mean, harmonic mean, variance, standard deviation, variance, dispersion, deviation, derivative, slope, spread, variance, range, skewness, kurtosis, L-moment, entropy, variance-to-mean ratio, maximum-to-minimum ratio, regularity, similarity, likelihood, correlation, covariance, autocorrelation, and autocovariance; or a sliding amount of any of the following functions of the rates: The system of claim 13.

15. The gait statistics include at least one of: presence of gait, interval, stride cycle time, step cycle time, N cycle time, interval distance, stride cycle distance, stride length, step cycle distance, step length, N cycle distance, asymmetry between even and odd cycles, asymmetry between modulo 1 and modulo 2 cycles, asymmetry between modulo 1 and modulo 3 cycles, asymmetry between modulo 1 and modulo 4 cycles, asymmetry between modulo 2 and modulo 3 cycles, asymmetry between modulo 2 and modulo 4 cycles, or asymmetry between modulo 3 and modulo 4 cycles. The system of claim 14.

16. The processor is further configured to detect a motion of the object when a magnitude of the motion statistics exceeds a first threshold or when an accumulation of the magnitudes of the motion statistics exceeds a second threshold. The system of claim 2.

17. The accumulation of the operational statistics includes at least one of each or any aggregation of: sum, integral, weighted sum, product, weighted product, average, weighted average, mean, arithmetic mean, geometric mean, harmonic mean, conditional mean, trimmed mean, statistic, median, weighted median, percentile, mode, maximum, minimum, variance, deviation, variance, sliding average, sliding mean, sliding weighted average, sliding median, sliding variance, sliding deviation, sliding statistic.

17. The system of claim 16.

18. 1. A method for wireless human-non-human motion detection, comprising: transmitting a wireless signal over a wireless multipath channel of the venue; receiving the wireless signal over the wireless multipath channel of the venue, the received wireless signal differing from the transmitted wireless signal due to the wireless multipath channel being affected by the motion of objects in the venue; obtaining a time sequence of channel information (TSCI) of the wireless multipath channel based on the received wireless signal; Calculating an autocorrelation function (ACF) based on the TSCI; and Calculating at least one ACF feature of the ACF or a function of the ACF; calculating an ACF statistic based on a first ACF feature of the ACF or the function of the ACF; calculating motion statistics associated with the motion of the object based on a second ACF feature of the ACF or the function of the ACF; calculating a velocity statistic related to a velocity of the movement of the object based on a third ACF feature of the ACF or the function of the ACF; calculating gait statistics associated with a walking motion of the object based on a fourth ACF feature of the ACF or the function of the ACF; In response to determining that the motion is detected, classifying the object associated with the detected motion as either human or non-human based on the ACF statistics, the motion statistics, the speed statistics, and the gait statistics. Including method.

19. calculating a time series of ACFs (TSACFs), each ACF being calculated based on a CI of a first sliding time window for each of the TSCIs; Calculating at least one said ACF feature of each ACF or said function of each ACF; calculating the ACF statistics based on the first ACF characteristics of each ACF or function of each ACF within a second sliding time window of the TSACF; Also includes 20. The method of claim 18.

20. calculating a time series of velocities (TSS) of the motion of the object, each of the velocities being calculated based on the third ACF features of each ACF or the function of each ACF in a third sliding time window of each of the TSACFs; calculating the speed statistics based on each speed of a fourth sliding time window of the TSS; Detecting the presence or absence of the walking motion of the object based on the TSS; calculating the gait statistics based on each speed of a fifth sliding time window of the TSS; Also includes 20. The method of claim 19.

Images