Spectral and spatial stitching for radio frequency sensing

Abstract

To provide a method and a system for communication and sensing systems.SOLUTION: The disclosed method includes the steps of: among other things, responsive to receiving a plurality of channel frequency response (CFR) measurements from an antenna operating at a low-bandwidth range, generating channel state information (CSI) across a high-bandwidth range based on the plurality of CFR measurements; and determining, based on the CSI, a Doppler shift across the high-bandwidth range. The system comprises a transmitter and a receiver coupled to the antenna, and a processing device coupled to the transmitter and the receiver.SELECTED DRAWING: None

Description

Technical Field

[0001] This disclosure relates to wireless devices, and more particularly to spectrum and spatial stitching for radio frequency sensing.

Background Art

[0002] Radio frequency (RF) sensing and localization for detecting and / or localizing individuals in space is useful in many applications, and in particular, can be used by wireless devices. Typically, wireless devices are configured by Wi-Fi(TM) technology or the IEEE 802.11 standard, but can also relate to wireless devices that communicate through other wireless local area network (WLAN) or personal area network (PAN) technologies including Bluetooth(R) (BT), Bluetooth(R) Low Energy (BLE), Zigbee(R), infrared, etc. In some cases, this type of wireless device can use PAN technology (e.g., Wi-Fi(TM)) as a radar device capable of detecting other objects including humans.

Summary of the Invention

Means for Solving the Problems

[0003] Aspects and embodiments of the present disclosure will be more fully understood from the detailed description given below and the accompanying drawings of various aspects and embodiments of the disclosure, which should not be regarded as limiting the disclosure to specific aspects or embodiments, but are given for explanation and understanding only.

Brief Description of the Drawings

[0004]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

DETAILED DESCRIPTION

[0005] Aspects of the present disclosure relate to spectral and spatial stitching for radio frequency (RF) sensing. A wireless device utilizes the emitted RF signals, collects information, and infers various insights without the need for dedicated sensors or additional infrastructure. When an RF signal is emitted, it interacts with objects, people, or the environment and can experience changes and distortions. These changes caused by reflection, diffraction, and absorption of the RF signal can be analyzed and interpreted to extract useful information. By monitoring the changes in the RF signal, the wireless device can perform detection and tracking of movement, presence, gestures, breathing patterns, object interactions, and other physical phenomena. The wireless device utilizes the channel state information (CSI) of the RF signal and / or other parameters, such as received signal strength (RSSI), to extract meaningful data. The CSI utilized by the wireless device can include noise such as high-bandwidth noise. Typically, the noise results from defects induced by various sources, such as phase-locked loop (PLL), phase offset, residual carrier frequency offset, and automatic gain control (AGC) operation.

[0006] The wireless device may be configured based on specific requirements for detection and localization. Selecting the number of antennas and / or the operating bandwidth of the antennas of the wireless device helps to meet the specific requirements. The operating bandwidth of the antennas of the wireless device determines the resolution and accuracy of detection and localization. For example, the higher the operating bandwidth (e.g., 80, 160, or 320 MHz), the finer the resolution, and thus the better detection and tracking of small movements and / or gestures are provided. On the other hand, the lower the operating bandwidth (e.g., 20 MHz), the coarser the resolution, and thus the detection and tracking of small movements and / or gestures are reduced. The number of antennas of the wireless device affects the spatial resolution of detection and localization. For example, if the wireless device has more antennas (e.g., four antennas), the wireless device can capture a wider view of the environment, and thus provides improved resolution and accuracy. For example, using large-scale CSI from antennas in different channels, detection and localization can be maximized. On the other hand, if the wireless device has only a few antennas (e.g., one antenna), the wireless device captures only one view of the environment, and thus limits the resolution of detection and localization.

[0007] Having more antennas in the wireless device and / or increasing the operating bandwidth of the antennas of the wireless device provides accuracy, resolution, and overall performance of detection and localization, but the wireless device can be costly and complex, increase power consumption, and affect battery life and overall energy efficiency.

[0008] Aspects and embodiments of the present disclosure address these and other limitations of existing technologies by enabling systems and methods that utilize wireless devices having few antennas and / or antennas operating at lower bandwidths, achieving the same performance as wireless devices having more antennas and / or antennas operating at higher bandwidths.

[0009] In particular, channel frequency response (CFR) measurements are obtained from different channels of an antenna (or multiple antennas at different times) at different times. As described above, the antenna operates in a low bandwidth range (e.g., 20 MHz). In some embodiments, CSI measurements or Doppler traces over a high bandwidth range (e.g., 80 MHz) are predicted from CFR measurements obtained in the low bandwidth range of the antenna.

[0010] The prediction of CSI measurements and / or Doppler traces over a high bandwidth range is based on the temporal coherence of the channel characteristics over time and the correlation between the spatial and frequency domains. Temporal coherence is related to the stability of the channel characteristics over time, indicating that the channel conditions remain relatively constant and change slowly within a specific duration. The correlation between the spatial and frequency domains is related to the interaction between spatial variations and frequency-selective channel effects. The spatial domain relates to the physical location, orientation, and movement of the device, and the frequency domain represents the distribution of signal power over different frequencies. Analyzing channel measurements in both domains enables a comprehensive understanding of how the channel response changes spatially and across different frequencies. In other words, due to temporal coherence and spatial and frequency domain correlations, CSI measurements and / or Doppler traces over a higher bandwidth can be predicted.

[0011] Aspects of the present disclosure overcome these drawbacks and others by predicting CSI or Doppler traces over a higher bandwidth range, thereby improving detection and localization performance using a limited frequency bandwidth and / or antennas.

[0012] FIG. 1 is a block diagram of an exemplary wireless device 100 configured for spectrum and spatial stitching for RF detection and localization, according to an example embodiment. In at least some embodiments, wireless device 100 includes, but is not limited to, a transmitter (TX) 102 (e.g., a PAN transmitter), a receiver (RX) 104 or RX (e.g., a PAN receiver), a communication interface 106, a transmitter (TX) antenna 110A coupled to transmitter 102, a receiver (RX) antenna 110B coupled to receiver 104, a memory 114, one or more input / output (I / O) devices 118 (e.g., a display screen, a touch screen, a keypad, etc.), and a processor 120. All of these components can be coupled to a communication bus 130. In some embodiments, aspects of communication interface 106 operate with processor 120 and operate as or perform the functions of a processing device of wireless device 100. In some embodiments, there is a single antenna and multiplexing logic that switches the use of the antenna between the transmitter and the receiver.

[0013] The RX antenna 110B may include a plurality of non-overlapping channels over a specific frequency band of the wireless device. For example, for the 2.4 GHz band, the plurality of channels are numbered from 1 to 14. Each of the plurality of channels has a specific center frequency and channel bandwidth. The center frequency represents the midpoint of the bandwidth range assigned to that channel. For example, for the 2.4 GHz band, the center frequency can be a value between 2.412 GHz and 2.484 GHz. The channel bandwidth indicates that the channel occupies a specific bandwidth range centered on the center frequency to which the channel is assigned. For example, for the 2.4 GHz band, the channel bandwidth can be from 20 MHz (e.g., low bandwidth range) to 320 MHz (e.g., high bandwidth range). Within the channel bandwidth, a plurality (e.g., 56) of subcarriers are assigned and are evenly spaced over the channel bandwidth. For example, for a 20 MHz channel bandwidth (i.e., a 20 MHz channel), the subcarriers are spaced at 312.5 kHz intervals.

[0014] In at least some embodiments, the memory 114 includes a storage device that stores instructions executable by the processor 120 and / or data generated by the communication interface 106. In various embodiments, front-end components, such as the transmitter 102, receiver 104, communication interface 106, and one or more antennas 110 described herein in various devices, are adapted or configured for WLAN and PAN-based frequency bands, for example, for Bluetooth® (BT), BLE, Wi-Fi™, Zigbee®, Z-wave™, etc.

[0015] In various embodiments, the communication interface 106 is incorporated as the front end of the wireless device 100, for example, into the transmitter 102 and the receiver 104. The communication interface 106 may cooperate to request / receive packets from other wireless devices or packets reflected from objects and / or humans when directed by the processor 120. The communication interface 106 can further process the data symbols received by the receiver 104 so that the processor 120 can perform further processing including object detection and position tracking. In some embodiments, the processing may process an RF signal spectrum or a heat map (e.g., a data array) to interpret return signal strength information (RSSI) including specific RF signals and / or channel characteristics of the communication link, referred to as channel state information (CSI). The CSI / RSSI data describes, for example, how an RF signal propagates from the transmitter 102 to the receiver 104 and represents, for example, the combined effects of scattering, fading, and power attenuation due to distance (representing the dynamic nature of the channel). The method by which CSI is measured is called channel estimation.

[0016] In these embodiments, the environment within the RF range of the wireless device 100 may contain a number of objects 50 that can include humans. The objects 50 can include human 50A as well as other objects 50B - 50N that are generally understood to be stationary objects. The RF range may vary depending on the type of WAN or PAN technology being used by the wireless device 100 at any given time. In some embodiments, the transmitter 102 emits an RF signal that causes one or more reflections from the objects 50, generating a reflected RF signal that is received by the receiver 104 and may include CSI / RSSI data.

[0017] In these embodiments, the wireless device 100 (also referred to herein as a detection RF wireless device) processes an array of data of reflected RF signals or active RF signals (also referred to as received RF signals), performs RF detection, detects the presence of one or more of the objects 50, and can determine the distance away from the wireless device 100 (e.g., the distance from the RX antenna 110B). Determining the distance away may be understood as localizing the object 50, the location of which is stored in the memory 114 and referenced at a later time, and may assist the disclosed algorithms and methods deployed to distinguish each of the objects 50 from one another, for example, to accurately determine the number and type of detected objects.

[0018] The processor 120 may further include a stitching component 122 and a machine learning (ML) model 124. The stitching component 122 with a single receiver may obtain channel frequency response measurements (e.g., CFR) from different channels of the antenna 110B with a low operating bandwidth (e.g., 20 MHz) at different times, and may generate a CFR or channel state information (CSI) over a high bandwidth range (e.g., 80, 160, or 320 MHz) for detection and localization, which will be described later. The stitching component 122 with multiple receivers, each having a low operating bandwidth (e.g., 20 MHz), may obtain CFRs from different receivers at different times, and may predict a CSI (or features extracted from the CSI, such as wavelet / time / frequency transforms, etc.) or a Doppler trace over a high bandwidth range (e.g., 80, 160, or 320 MHz), which will be described later. Depending on the embodiment, the processor 120 may include a machine learning (ML) model 124 and may assist in generating a CFR, CSI, or Doppler trace over a high bandwidth range.

[0019] FIG. 2 is a simplified block diagram 200 of a stitching component (e.g., stitching component 122 of FIG. 1) of a wireless device according to an embodiment of the present disclosure. The stitching component 122 may include a non-uniform discrete Fourier transform (NDFT) 210, a prediction filter 220, a phase sanitizer 230, and a feature extractor 240.

[0020] The stitching component 122 may receive a plurality of CFRs. Each CFR of the plurality of CFRs is a frequency domain representation of the characteristics of a channel indicating how the channel affects the transmitted RF signal over different frequencies. Each CFR associated with a channel is a set of subcarrier CFRs. A subcarrier CFR is a CFR obtained from the subcarriers of the channel. The plurality of CFRs are obtained from different channels of the plurality of channels of the antenna 110B of FIG. 1 at different times. The channel bandwidth of the plurality of channels may be 20 MHz (e.g., low bandwidth range). In other words, the plurality of CFRs are obtained from a low bandwidth range (e.g., 20 MHz) over a plurality of carrier frequencies.

[0021] In one embodiment, the stitching component 122 may select a plurality of sub-carrier CFRs from a plurality of CFRs. The stitching component 122 selects a subset of the set of sub-carrier CFRs associated with each CFR from each of the plurality of CFRs. The stitching component 122 determines the number of sub-carrier CFRs from the set of sub-carrier CFRs associated with each CFR to be selected based on the age of each CFR. For example, the newer each CFR is, the more sub-carrier CFRs are selected from the set of sub-carrier CFRs associated with each CFR, and the older each CFR is, the fewer sub-carrier CFRs are selected from the set of sub-carrier CFRs associated with each CFR. In some embodiments, the stitching component 122 may use a weight value, which is applied to the number of sub-carrier CFRs in the set of sub-carrier CFRs to assist in the selection. The stitching component 122 may provide the plurality of selected sub-carrier CFRs to the NDFT210 of the stitching component 122.

[0022] The NDFT210 is configured to generate CFRs over a higher bandwidth range from CFRs sampled non-uniformly in the frequency domain. In particular, the NDFT210 reconstructs a time-domain RF signal from the non-uniformly sampled CFRs. The NDFT210 may estimate values between and outside the non-uniformly sampled CFRs using interpolation and / or extrapolation. Thus, the NDFT210 may provide CFRs over a higher bandwidth range (e.g., 80, 160, or 320 MHz).

[0023] The stitching component 122 may derive CSI over a high bandwidth range from CFR over a high bandwidth range. CSI provides information used to characterize composite effects such as channel gain, phase shift, frequency response, noise level, path loss, scattering, diffraction, fading, shadowing, etc. when an RF signal propagates from a transmitter to a receiver. The stitching component 122 may generate a Doppler trace that represents the Doppler effect (or Doppler shift) observed from CSI over time based on the CSI. The Doppler effect occurs when there is relative motion between the transmitter and the receiver of a wireless device. The Doppler effect typically introduces changes in CSI due to changes in an object or human target that assist in detecting motion. The Doppler trace can reveal the velocity, direction, and time-dependent characteristics of the channel.

[0024] In other embodiments, the stitching component 122 may provide a plurality of CFRs to the prediction filter 220 of the stitching component 122. A prediction filter 220, such as a Kalman prediction filter, is a recursive algorithm used to estimate and / or predict future values of an RF signal based on past observations or measurements (e.g., a plurality of CFRs). In particular, a low bandwidth (e.g., 20 MHz) includes a predetermined number of subcarriers (e.g., 64 subcarriers) associated with the low bandwidth, and similarly, a higher bandwidth (e.g., 80 MHz) includes a predetermined number of subcarriers (e.g., 256 subcarriers) associated with the higher bandwidth. The prediction filter 220 includes a state of a predetermined number of dimensions (e.g., 512 dimensions) that can be estimated. In some embodiments, each dimension state of the state of a predetermined number of dimensions is represented by [CSI, ΔCSI], and CSI represents a complex value over each subcarrier over a predetermined number of subcarriers (e.g., 256 subcarriers) associated with the higher bandwidth.

[0025] Accordingly, at a given time, in the prediction filter 220, the CSI of a predetermined number of subcarriers (e.g., 64 subcarriers) associated with a low bandwidth is measured. The prediction filter 220 implements a constant velocity model and models how the CSI of each of the remaining subcarriers (e.g., 256 subcarriers - 64 subcarriers) changes over time. The uncertainty matrix is initialized by a multiplier of the identity matrix (e.g., 0.2) and learns confidence over time. The gain of the prediction filter 220 helps maintain a balance between the measurement and the prediction based on the confidence determined by the corresponding entry of the uncertainty matrix. The prediction filter 220 utilizes a state space model to capture the global relationships of the changing subcarriers and their local relationships from the neighborhood of those subcarriers.

[0026] In other embodiments, the stitching component 122 may provide a plurality of CFRs to a machine learning model (e.g., a single antenna ML model 250A) of the ML model 124 of FIG. 1. The single antenna ML model 250A may be configured to generate CSI over a high bandwidth range based on the plurality of CFRs. The single antenna ML model 250A may include a first fully connected layer (FCN), a long short-term memory (LSTM), and a last FCN.

[0027] FCN means an architecture in which all neurons within a layer are connected to all neurons in the previous layer. Each neuron, also referred to as a perceptron, unit or node, is a mathematical function that collects and classifies information. In some cases, a neuron takes weight values, performs mathematical calculations, and generates an output. Each neuron receives inputs from all neurons in the previous layer and generates an output value based on the weighted sum of these inputs, which represent the parameters of the model learned during the training process. The output of each neuron in a fully connected layer typically passes through an activation function, such as a sigmoid, rectified linear unit (ReLU) or hyperbolic tangent function, to mitigate non-linearity. Typically, the FCN used as the last few layers of an ML model converts high-level features learned in the previous layer into meaningful outputs.

[0028] LSTM is a type of recurrent neural network (RNN) designed to capture long-term dependencies in sequential data over time. LSTM provides memory cells within the network that selectively store, access, and update information over time. A memory cell consists of four main components, namely, a cell state (ct), an input gate (i), a forget gate (f), and an output gate (o). The cell state acts as long-term memory and operates throughout the entire sequence. The input gate controls the flow of new information into the cell state and determines which information is to be stored. The forget gate determines which information should be discarded or forgotten from the cell state. The output gate regulates the flow of information from the cell state to the output or the next hidden state. By selectively storing and updating information within the memory cell, LSTM can effectively capture and remember long-term dependencies in sequential data over time. LSTM can be trained to predict or filter future observations based on past data, and the past data generates a sequence of predicted values that represent future points in a time series.

[0029] In response to receiving a plurality of CFRs by the single antenna ML model 250A, the first FCN of the single antenna ML model 250A assists in mitigating the non-linearity of the plurality of CFRs. The output of the first FCN is transferred to the LSTM of the single antenna ML model 250A to track and filter a plurality of CFRs over a higher bandwidth range (e.g., 80, 160, or 320 MHz). The output of the LSTM is transferred to the final FCN, and the final FCN converts a sequence of predicted values associated with the output of the LSTM into the constituent frequencies that form the RF signal. Depending on the embodiment, the final FCN may be replaced by a discrete Fourier transform (DFT) to obtain the constituent frequencies that form the RF signal.

[0030] The stitching component 122 may derive CSI over a high bandwidth range from the CFR over the high bandwidth range. The stitching component 122 may use various techniques, such as phase connection methods, reference signal calibration, interpolation or filtering, or channel tracking, e.g., performing phase sanitization, to remove and / or mitigate phase ambiguities or inconsistencies in the CSI. The stitching component 122 may further generate a Doppler trace based on the CSI.

[0031] In other embodiments, the stitching component 122 may derive a CSI associated with each CFR for each of the plurality of CFRs. The stitching component 122 may perform phase sanitization for each of the derived CSIs. The stitching component 122 may further generate Doppler traces aggregated into a plurality of Doppler traces for each of the derived CSIs. The stitching component 122 may provide the plurality of Doppler traces to a single antenna ML model 250A. In response to receiving the plurality of Doppler traces by the single antenna ML model 250A, the first FCN of the single antenna ML model 250A assists in mitigating the non-linearity of the plurality of Doppler traces. The output of the first FCN is transferred to the LSTM of the single antenna ML model 250A to track and filter the plurality of Doppler traces over a higher bandwidth range (e.g., 80, 160, or 320 MHz). The output of the LSTM is transferred to a final FCN, and the final FCN converts a sequence of predicted values associated with the output of the LSTM to the constituent frequencies forming the RF signal. Depending on the embodiment, the final FCN may be replaced by a DFT to obtain the constituent frequencies forming the RF signal. As a result, the RF signal may be used for further tasks such as presence detection, activity detection, geofencing activity detection, location determination, etc.

[0032] As described above, the wireless device may include a plurality of receivers. Each of the plurality of receivers may operate at 20 MHz (e.g., a low bandwidth range). The stitching component 122 may receive CFRs from different receivers of the plurality of receivers at different times. The stitching component 122 may derive a CSI from the CFR associated with each receiver for each of the plurality of receivers.

[0033] In one embodiment, the stitching component 122 may provide, for each of a plurality of receivers, the CSI associated with each receiver to an ML model (e.g., the multi-antenna ML model 250B) of the ML model 124. That is, for each receiver of the wireless device, the corresponding multi-antenna ML model 250B receives the CSI associated with each receiver. Each multi-antenna ML model 250B may be an LSTM. The multi-antenna ML model 250B may predict a portion of the CSI over a higher bandwidth range (e.g., 80, 160, or 320 MHz) based on the CSI associated with each receiver. Thus, the outputs of each of the multi-antenna ML models 250B associated with the receivers can be combined to provide the predicted CSI over a higher bandwidth range.

[0034] In other embodiments, the stitching component 122 may perform phase sanitization for each of a plurality of receivers and / or generate Doppler traces or other features extracted from the CSI associated with each receiver. The stitching component 122 may provide, for each of a plurality of receivers, the Doppler trace associated with each receiver to an ML model (e.g., the multi-antenna ML model 250B) of the ML model 124. Each multi-antenna ML model 250B may be an LSTM. The multi-antenna ML model 250B may predict a portion of the Doppler trace over a higher bandwidth range (e.g., 80, 160, or 320 MHz) based on the Doppler trace associated with each receiver. Thus, the outputs of each of the multi-antenna ML models 250B associated with the receivers can be combined to provide the predicted Doppler trace over a higher bandwidth range. In other words, a plurality of CFRs from a low bandwidth range (e.g., 20 MHz) over a plurality of carrier frequencies are stitched (using the NDFT 210 or the prediction filter 220) to obtain a representative CFR over a larger bandwidth.

[0035] FIG. 3 is a flowchart of a method 300 for predicting Doppler traces or other features extracted from CSI over a higher bandwidth range for a wireless device having a single antenna, according to an embodiment of the present disclosure. The method 300 is executable by processing logic that can include hardware (e.g., a processing device, circuitry, dedicated logic, programmable logic, microcode, the hardware of a device, an integrated circuit, etc.), software (e.g., instructions that operate or execute on a processing device), or a combination thereof. In some embodiments, the method 300 is executed by a wireless device 100 that includes a stitching component 122 and / or a processor 120 (e.g., a processing device).

[0036] In operation 310, the processing logic receives CFR measurements from an antenna operating in a low bandwidth range.

[0037] In operation 320, the processing logic generates CSI over a high bandwidth range based on the CFR measurements. In response to receiving CFR measurements (e.g., a plurality of channel frequency response (CFR) measurements) from an antenna operating in a low bandwidth range, the processing logic generates channel state information (CSI) over a high bandwidth range based on the plurality of CFR measurements. Each CFR measurement of the plurality of CFR measurements may be measured from different channels of the antenna at different times.

[0038] In one embodiment, to generate CSI over a high bandwidth range, the processing logic identifies a subset of the plurality of CFR measurements. The processing logic obtains one or more subcarrier CFR measurements out of the plurality of subcarrier CFR measurements of each CFR measurement of the subset. In some embodiments, to obtain one or more subcarrier CFR measurements, the processing logic determines the age of each CFR measurement of the subset for each CFR measurement of the subset.

[0039] The processing logic identifies, based on the age, the number of subcarrier CFR measurements to select from a plurality of subcarrier CFR measurements of each CFR measurement. To identify, based on the age, the number of subcarrier CFR measurements to select from a plurality of subcarrier CFR measurements of each CFR measurement, the processing logic determines the number of subcarrier CFR measurements associated with the plurality of subcarrier CFR measurements of each CFR measurement.

[0040] The processing logic applies, based on the age, a weight value to the number of subcarrier CFR measurements and determines the number of subcarrier CFR measurements to select from a plurality of subcarrier CFR measurements of each CFR measurement. The processing logic selects one or more subcarrier CFR measurements based on the identified number. As described above, the newer each CFR is, more subcarrier CFRs in the set of subcarrier CFRs associated with each CFR may be selected, and the older each CFR is, fewer subcarrier CFRs in the set of subcarrier CFRs associated with each CFR may be selected. In some embodiments, the CFR is determinable based on the overall measured SNR / RSSI.

[0041] The processing logic converts one or more subcarrier CFR measurements associated with each CFR measurement of the subset into CSI over a high bandwidth range. As described above, to convert one or more subcarrier CFR measurements associated with each CFR measurement of the subset into CSI over a high bandwidth range, the processing logic provides the one or more subcarrier CFR measurements to an NDFT, which is configured to generate a CFR over a higher bandwidth range from a CFR sampled non-uniformly in the frequency domain.

[0042] In other embodiments, to generate CSI over a high bandwidth range, the processing logic provides each CFR measurement of a plurality of CFR measurements to a prediction filter. As described above, the prediction filter may be a recursive algorithm used to estimate and / or predict future values of the RF signal based on a plurality of CFR measurements. Accordingly, the processing logic obtains a plurality of predicted CSI over a high bandwidth range from the prediction filter, combines the predicted CSI over the high bandwidth range, and generates a plurality of measured CSI over the high bandwidth range.

[0043] In yet other embodiments, to generate CSI over a high bandwidth range, the processing logic provides a plurality of CFR measurements to a machine learning model and obtains CSI over the high bandwidth range from the machine learning (ML) model (e.g., a single antenna ML model). As described above, the ML model may mitigate the non-linearity of a plurality of CFRs, and then may track and filter a plurality of CFRs over a higher bandwidth range and may convert the sequence of predicted values to the constituent frequencies that form the RF signal.

[0044] In operation 330, the processing logic determines a Doppler shift based on the CSI. In other words, the processing logic determines a Doppler shift over a high bandwidth range based on the CSI. The high bandwidth range (e.g., 80 MHz) may be larger than the low bandwidth range (e.g., 20 MHz).

[0045] FIG. 4 is a flowchart of a method 400 for predicting Doppler traces over a higher bandwidth range for a wireless device having a single antenna, according to an embodiment of the present disclosure. Method 400 is executable by processing logic that can include hardware (e.g., a processing device, circuitry, dedicated logic, programmable logic, microcode, hardware of a device, an integrated circuit, etc.), software (e.g., instructions that operate or execute on a processing device), or a combination thereof. In some embodiments, method 400 is executed by a wireless device 100 that includes a stitching component 122 and / or a processor 120 (e.g., a processing device).

[0046] In operation 410, the processing logic receives CFR measurements from an antenna operating in a low bandwidth range. In other words, a plurality of channel frequency response (CFR) measurements are received from an antenna operating in a low bandwidth range.

[0047] In operation 420, the processing logic generates a plurality of Doppler traces over the low bandwidth range based on the CFR measurements. In other words, for each CFR measurement of the plurality of CFR measurements, a first Doppler trace over the low bandwidth range is generated. As described above, to generate the first Doppler trace over the low bandwidth range, the processing logic determines the channel state information (CSI) of the antenna based on the plurality of CFR measurements and determines the first Doppler trace over the low bandwidth range based on the CSI of the antenna.

[0048] In operation 430, the processing logic determines a Doppler shift over a high bandwidth range based on a plurality of Doppler traces over a low bandwidth range. In other words, a second Doppler shift over a high bandwidth range is generated based on a first Doppler trace associated with each CFR measurement of a plurality of CFR measurements. To generate the second Doppler trace, the processing logic provides the first Doppler trace to a machine learning model and obtains the second Doppler trace. The high bandwidth range (e.g., 80 MHz) may be larger than the low bandwidth range (e.g., 20 MHz).

[0049] FIG. 5 is a flow diagram of a method 500 for predicting channel characteristic evaluation for a wireless device having a plurality of antennas, according to an embodiment of the present disclosure. Method 500 is executable by processing logic that can include hardware (e.g., a processing device, circuitry, dedicated logic, programmable logic, microcode, hardware of a device, an integrated circuit, etc.), software (e.g., instructions that operate or execute on a processing device), or a combination thereof. In some embodiments, method 500 is executed by a wireless device 100 that includes a stitching component 122 and / or a processor 120 (e.g., a processing device).

[0050] In operation 510, the processing logic receives CFR measurements from an antenna operating at a first frequency. In other words, CFR measurements are received from each of a plurality of antennas of a wireless device operating in a low bandwidth range. The CFR measurements associated with each of the plurality of antennas may be measured at different times. For example, the CFR measurement for a first subcarrier is obtained at antenna 1 at time T, at antenna 2 at time T + 1, at antenna 3 at time T + 2, at antenna 4 at T + 3, and so on. In some embodiments, the measurements are obtained from each antenna by switching between antennas (e.g., via I / O pins).

[0051] In operation 520, the processing logic presents the CFR measurements to the ML model and obtains an evaluation of channel characteristics over a high bandwidth range. In other words, for each CFR measurement associated with one of the plurality of antennas, the data associated with each CFR measurement is provided to one of the plurality of ML models. The data associated with each CFR measurement may be CSI measurements or Doppler traces derived from the CFR measurements. Each of the plurality of machine learning models may be a long short-term memory (LSTM).

[0052] The output of each of the plurality of ML models is an evaluation of channel characteristics over a portion of the high bandwidth range. To determine the evaluation of channel characteristics over the high bandwidth range, the processing logic combines the output of each of the plurality of machine learning models to generate an evaluation of channel characteristics. The high bandwidth range (e.g., 80 MHz) is larger than the low bandwidth range (e.g., 20 MHz). The evaluation of channel characteristics may be CSI or a Doppler trace. For example, the evaluation of channel characteristics is obtained for antennas 1 to 4 at T + 3 using all past CSI (or CFR) measurements from T to T + 3. Depending on the embodiment, a plurality of ML models can be used to reconstruct CSI (or CFR) measurements or Doppler traces over a higher bandwidth range based on CSI (or CFR) measurements over different subcarriers, and / or reconstruct CSI (or CFR) measurements or Doppler traces over multiple antennas based on CSI (or CFR) measurements over different antennas.

[0053] Throughout this specification, references to "one embodiment", "an embodiment", "embodiments" or "the embodiment" mean that a particular feature, structure, or characteristic described in connection with the embodiment(s) and / or embodiment(s) is included in at least one embodiment and / or embodiment. Thus, the appearances of the phrases "in one embodiment" or "in an embodiment" in various places throughout this specification are not necessarily referring to the same embodiment, but may do so depending on the context. Further, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0054] The terms "include", "including", "contain" and their variants and other similar terms, as used in the detailed description or claims, are intended to be inclusive in a manner similar to the term "comprise" as an open transitional term without excluding any additional or other elements.

[0055] As used herein, terms such as "component", "module", "system", etc. are generally intended to mean an entity related to a computer-related entity, hardware (e.g., circuitry), software, a combination of hardware and software, or an operable machine having one or more specific functions. For example, a component can be, but is not limited to, a process operating on a processor (e.g., a digital signal processor), a processor, an object, an executable file, a thread of execution, a program, and / or a computer. By way of example, both an application operating on a controller and the controller can be components. One or more components can exist within a process and / or a thread of execution, a component can be localized on one computer, and / or can be distributed between two or more computers. Further, a "device" can be in the form of specially designed hardware, general-purpose hardware specialized by execution of software that enables the hardware to perform a specific function (e.g., generation of a point of interest and / or descriptor), software on a computer-readable medium, or a combination thereof.

[0056] The systems, circuits, modules, etc. described above are described in terms of the interactions between some components and / or blocks. It should be recognized that this type of system, circuit, component, block, etc. can include those components or some of the designated sub-components, some of the designated components or sub-components, and / or additional components, and follow various permutations and combinations of those described above. The sub-components can also be implemented as components communicatively coupled to other components, rather than being included (hierarchically) within a parent component. Additionally, it should be noted that one or more components can be combined to form a single component that provides an aggregated performance, or can be divided into several separate sub-components, and one or more central layers such as a management layer can be provided, communicatively coupled to and providing an integrated function for this type of sub-component. Any component described in this specification can also interact with one or more other components that are not specifically described in this specification but are known to those skilled in the art.

[0057] Furthermore, in this specification, the terms "example" or "exemplary" are used to mean functioning as an example (example, instance, illustration). In this specification, any aspect or design described as "exemplary" should not necessarily be construed as being more preferred or advantageous than other aspects or designs. Rather, the use of the terms "example" or "exemplary" is intended to present the concept in a specific way. When used in this application, the term "or" is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless otherwise specified or clear from the context, "X uses A or B" is intended to mean any of the natural inclusive permutations. That is, "X uses A or B" is satisfied under any of the above examples when X uses A, X uses B, or X uses both A and B. In addition, the indefinite article in the singular used in this application and the appended claims should generally be construed to mean "one or more" unless otherwise specified or clear from the context.

[0058] Finally, the embodiments described in this specification include the collection of data that describes a user and / or the user's activities. In one embodiment, this type of data is collected only when the user consents to the collection of this data. In some embodiments, the user is prompted to explicitly permit the data collection. Furthermore, the user may opt-in or opt-out when participating in this type of data collection activity. In one embodiment, the collected data is anonymized before any analysis is performed and any statistical patterns are obtained so that the user's identity cannot be determined from the collected data.

Claims

1. a transmitter coupled to at least one antenna; a receiver coupled to the at least one antenna; and a processing device coupled to the transmitter and the receiver; A device comprising: The processing device includes: in response to receiving a plurality of channel frequency response (CFR) measurements from an antenna operating in a low bandwidth range, generating channel state information (CSI) across a high bandwidth range based on the plurality of CFR measurements; determining a Doppler shift over the wide bandwidth range based on the CSI; and performing operations including device.

2. Each CFR measurement of the plurality of CFR measurements is measured from a different channel of the antenna at various times. The device of claim 1 .

3. the high bandwidth range is greater than the low bandwidth range; The device of claim 1 .

4. Generating the CSI across the wide bandwidth range includes: identifying a subset of the plurality of CFR measurements; obtaining, from each CFR measurement value of the subset, one or more subcarrier CFR measurements from a plurality of subcarrier CFR measurement values ​​for the respective CFR measurement value; converting the one or more subcarrier CFR measurements associated with each CFR measurement of the subset to the CSI spanning the wide bandwidth range; Including, The device of claim 1 .

5. Obtaining one or more subcarrier CFR measurements from each CFR measurement of the subset includes: determining, for each CFR measurement of the subset, an age of the respective CFR measurement; identifying a number of subcarrier CFR measurements to select from the plurality of subcarrier CFR measurements for the respective CFR measurement based on the age; selecting the one or more subcarrier CFR measurements based on the identified number; and Including, The device of claim 4.

6. Identifying the number of subcarrier CFR measurements to select from the plurality of subcarrier CFR measurements for the respective CFR measurements based on the age includes: determining a number of subcarrier CFR measurements associated with the plurality of subcarrier CFR measurements for the respective CFR measurements; applying a weighting value to the number of subcarrier CFR measurements based on the age to determine the number of subcarrier CFR measurements to select from the plurality of subcarrier CFR measurements for the respective CFR measurements; Including, The device of claim 5.

7. Generating the CSI across the wide bandwidth range includes: providing each CFR measurement of the plurality of CFR measurements to a predictive filter; obtaining a plurality of predicted CSIs spanning the wide bandwidth range from the prediction filter; combining the multiple predicted CSIs across the wide bandwidth range to generate the CSI across the wide bandwidth range; and Including, The device of claim 1 .

8. Generating the CSI across the wide bandwidth range includes: providing the plurality of CFR measurements to a machine learning model; and obtaining the CSI across the high bandwidth range from the machine learning model; and Including, The device of claim 1 .

9. in response to receiving a plurality of channel frequency response (CFR) measurements from an antenna operating in a low bandwidth range, generating, for each CFR measurement of the plurality of CFR measurements, a first Doppler trace spanning the low bandwidth range; generating a second Doppler shift over a high bandwidth range based on the first Doppler trace associated with each CFR measurement of the plurality of CFR measurements; The method includes:

10. The step of generating the first Doppler trace over the low bandwidth range comprises: determining channel state information (CSI) for the antenna based on the plurality of CFR measurements; determining the first Doppler trace across the low bandwidth range based on the CSI of the antenna; Including, 10. The method of claim 9.

11. The step of generating a second Doppler trace based on the first Doppler trace comprises: providing the first Doppler trace to a machine learning model; obtaining the second Doppler trace from the machine learning model; Including, 10. The method of claim 9.

12. the high bandwidth range is greater than the low bandwidth range; 10. The method of claim 9.

13. Each CFR measurement of the plurality of CFR measurements is measured from a different channel of the antenna at various times.

10. The method of claim 9.

14. receiving a channel frequency response (CFR) measurement from each antenna of a plurality of antennas of a wireless device operating in a low bandwidth range; for each CFR measurement associated with an antenna of the plurality of antennas, providing data associated with the respective CFR measurement to a machine learning model of a plurality of machine learning models; obtaining an output from each machine learning model of the plurality of machine learning models; determining a channel characterization over a wide bandwidth range based on the output of each machine learning model of the plurality of machine learning models; The method includes:

15. the CFR measurements associated with each antenna of the plurality of antennas are measured at different times; The method of claim 14.

16. Each of the plurality of machine learning models is a long short-term memory (LSTM). The method of claim 14.

17. determining the channel characterization over a wide bandwidth range based on the output of each machine learning model of the plurality of machine learning models, combining the outputs of each machine learning model of the plurality of machine learning models to generate the channel characterization, each output representing a portion of the channel characterization. The method of claim 14.

18. the high bandwidth range is greater than the low bandwidth range; The method of claim 14.

19. the channel characterization is one of CSI or Doppler trace; The method of claim 14.

20. the data associated with each CFR measurement is one of a CSI measurement or a Doppler trace derived from the CFR measurement. The method of claim 14.

Images