Fully local wireless sensing device, method for estimating potential living organisms, and program

Abstract

This system enables monitoring of potential human and animal targets within a surveillance space or through obstacles, without using cameras or relying on cloud communication. [Solution] CSI between transmitting and receiving wireless devices is acquired locally, and the difference from the vacant room standard, respiratory cycle, movement occupancy rate, subcarrier deviation distribution, etc. are integrated to estimate the number of human candidates, representative respiratory peaks, and animal candidates. The estimation results are displayed in a three-dimensional pseudo-display, local AI comments are generated, and in the event of danger, a summary notification is sent via email or mobile phone notification as needed.

Description

Technical Field

[0001] The present invention relates to a complete local wireless sensing device, method, and program using channel state information of wireless communication signals, and particularly to a monitoring technology for three-dimensionally displaying person candidates, multiple person candidates, representative respiratory peaks, and animal candidates without using a camera and without transmitting raw data to an external cloud.

Background Art

[0002] In wireless communication such as Wi-Fi, channel state information (hereinafter referred to as CSI) indicating the propagation path state between transmission and reception may be obtained. Since CSI can reflect minute multipath changes associated with human movement, stillness, respiration, etc., its application to presence detection, motion detection, and respiration estimation without using a camera has been studied.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

[0005] While conventional technologies demonstrate the ability to detect respiration or individuals from CSI, they do not adequately disclose an integrated configuration that simultaneously handles candidate individuals and animal candidates, distinguishes representative respiration peaks from individual candidates, displays three-dimensional pseudo-skeletons and animal candidates on the same local screen, and generates local AI comments without relying on the internet.

[0006] The objective of this invention is to provide a three-dimensional display and local AI comments that are less likely to cause misunderstandings for the user, by locally estimating the number of potential human candidates, representative respiratory peaks, and potential animal candidates based on wireless CSI, without relying on cameras, wearables, or external clouds. [Means for solving the problem]

[0007] According to one aspect of the present invention, a fully local wireless sensing device is provided, comprising: an acquisition unit that acquires channel state information relating to a wireless signal transmitted from a transmitting wireless device and received by a receiving wireless device; a formatting unit that formats the channel state information as time-series subcarrier information; a difference calculation unit that calculates the difference between vacant room reference information based on a reference state and the current subcarrier information; a respiratory estimation unit that extracts periodic components of the respiratory frequency band from the subcarrier information and determines a representative respiratory peak; and a living organism estimation unit that estimates the number of human candidates and animal candidates based on the difference, the periodic components, a motion index, and a subcarrier deviation distribution.

[0008] The fully local wireless sensing device may include a display control unit that generates a display of human candidates corresponding to the number of human candidates and a display of animal candidates corresponding to the number of animal candidates in the same three-dimensional display space. The display control unit may display a transmitting wireless device, a receiving wireless device, a propagation path, a voxel point cloud, a pseudo-skeleton of human candidates, a pseudo-quadrupedal display of animal candidates, and a representative index indicating the assignment destination of the representative respiratory peak in the three-dimensional display space.

[0009] The respiration estimation unit may analyze the subcarrier information within a predetermined time window and determine the representative respiration peak by frequency spectrum analysis, autocorrelation analysis, principal component analysis, or a combination thereof in the respiration frequency band. The representative respiration peak is a representative value obtained from the reflected components mixed throughout the entire monitoring space and does not necessarily represent the individual respiration rate of a specific person or animal; therefore, it may be displayed separately from individual candidates.

[0010] The organism estimation unit may estimate the number of candidate people as a step value of one, two, or three or more, using at least one of the persistence of multiple candidate respiratory peaks, the occupancy rate of the movement index, the upper quantile intensity of the movement index, and the spread of the subcarrier deviation distribution. Alternatively, the organism estimation unit may estimate a candidate animal different from the candidate person display based on at least one of the difference, the movement index, the reflex characteristics corresponding to body height, the time variation characteristics corresponding to gait, or the subcarrier deviation characteristics corresponding to surface scattering.

[0011] The animal candidates are not limited to dogs, but may include cats, livestock, service animals, exhibition animals, pets, and other animals. The animal candidate display is not limited to a display that confirms that it is a specific species of animal, but may also be a display that shows users a living organism or animal reflector that has different reflective or movement characteristics from the human candidate.

[0012] The fully local wireless sensing device may include a control unit that operates the acquisition unit, the shaping unit, the difference calculation unit, the respiration estimation unit, the living organism estimation unit, and the display control unit locally without transmitting the raw channel status information to an external cloud. This allows monitoring of human candidates, multiple human candidates, representative respiration peaks, and animal candidates in a configuration that does not transmit camera images or raw channel status information to an external cloud.

[0013] The control unit may generate natural language comments using a local inference model or local rules, taking at least one of the number of person candidates, the animal candidates, the representative respiratory peak, respiratory quality, movement index, vacancy difference, and risk level as input. The natural language comments may include the state of the monitoring space, the confidence level of the respiratory estimation, the number of person candidates, the presence or absence of animal candidates, whether notification is required, or confirmation items for the user.

[0014] Furthermore, in one aspect of the present invention, the system may include a risk determination unit and a notification control unit for determining the degree of risk. The notification control unit may suppress notifications when the degree of risk is normal or under observation, and may only send a summary notification to an external party when the degree of risk is determined to be dangerous.

[0015] The summary notification may include at least one of the following: risk level, status label, time, number of potential people, potential animals, representative respiratory peak, respiratory quality, local AI comment, and response recommendation. The summary notification may be sent via email, SMS, MMS, push notification to a mobile phone terminal, or notification service corresponding to a mobile phone number or email address. The summary notification may not include the raw data of channel status information, the I / Q column, or the subcarrier column itself.

[0016] According to another aspect of the present invention, a fully local wireless sensing method is provided, which includes the steps of: locally acquiring channel state information relating to a wireless signal transmitted from a transmitting wireless device and received by a receiving wireless device; formatting the channel state information as time-series subcarrier information; calculating the difference between vacant room reference information based on a reference state and the current subcarrier information; extracting periodic components of the respiratory frequency band from the subcarrier information to determine a representative respiratory peak; estimating the number of human candidates and animal candidates based on the difference, the periodic components, a motion index, and a subcarrier deviation distribution; and generating a human candidate display corresponding to the number of human candidates and an animal candidate display corresponding to the animal candidates in the same three-dimensional display space, and displaying the representative respiratory peak separately from individual candidates.

[0017] According to yet another aspect of the present invention, a program is provided for a computer that enables the following functions: formatting channel state information relating to radio signals between transmitting and receiving wireless devices as time-series subcarrier information; calculating the difference between vacant room reference information based on a reference state and the current subcarrier information; extracting periodic components of the respiratory frequency band from the subcarrier information to determine a representative respiratory peak; estimating the number of human candidates and animal candidates based on the difference, the periodic components, a motion index, and a subcarrier deviation distribution; and generating displays of human candidates corresponding to the number of human candidates and displays of animal candidates corresponding to the animal candidates in the same three-dimensional display space, and displaying the representative respiratory peak separately from individual candidates.

[0018] The apparatus, method, or program described above is not limited to a single transmit / receive link, but may be implemented using a network of multiple transmitting radio devices, multiple receiving radio devices, multiple frequency bands, multiple antennas, or multiple rooms. This can improve the accuracy of separating candidate people, candidate animals, candidate breathing, and candidate locations. [Effects of the Invention]

[0019] According to the present invention, it becomes easier to visually identify potential individuals, multiple potential individuals, and potential animals in the same space while enhancing privacy without the need for cameras. According to the present invention, it is possible to clearly display on the screen that the representative respiratory rate is the representative peak of the entire space, and suppress misidentification with the determined respiratory rate of an individual person. According to the present invention, since the raw CSI data is not transmitted to an external cloud, it is easy to continue monitoring even in a place with a poor communication environment, a place where external communication is restricted, or a place where privacy requirements are high.

[0020] According to the present invention, excessive external notifications are suppressed during normal times or at the attention stage, and when the degree of danger is determined to be "dangerous", summary information can be promptly notified to protectors, administrators, drivers, work managers, breeders, and other notification destinations by email or mobile phone notifications.

Brief Description of the Drawings

[0021] [Figure 1] It is a diagram showing the overall configuration of a system according to an embodiment of the present invention. [Figure 2] It is a functional block diagram of a local processing device. [Figure 3] It is a diagram showing a processing flow. [Figure 4] It is a diagram showing a three-dimensional virtual display screen. [Figure 5] It is a diagram showing respiration estimation and representative peak separation. [Figure 6] It is a diagram showing the estimation of the number of people and the estimation of animal candidates. [Figure 7] It is a diagram showing a complete local AI comment, danger determination, and external notification for email or mobile phone. [Figure 8] It is a diagram showing an embodiment and an application field.

Modes for Carrying Out the Invention

[0022] Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the present invention is not limited to the following embodiments, and each component can be replaced, added, omitted, or combined without departing from the spirit of the present invention. [[ID=^]]

[0023] As shown in Figure 1, the fully local wireless sensing device 1 includes a transmitting wireless device 10, a receiving wireless device 20, a monitoring space 40, a local processing unit 50, a display unit 60, and a notification unit 70. The transmitting wireless device 10 and the receiving wireless device 20 are low-power wireless devices with Wi-Fi functionality, such as ESP32-S3. The local processing unit 50 is an edge computer, such as a Raspberry Pi.

[0024] In this specification, the term "monitoring space" or "detection target area" is not limited to the same room in which the transmitting wireless device 10 and the receiving wireless device 20 are located. The detection target area may also include adjacent or separate spaces separated by walls, doors, furniture, partitions, vehicle body components, building components, temporary components, or other barriers.

[0025] The propagation state of wireless signals can change depending on the material, thickness, moisture content, frequency band, transceiver arrangement, and transmission power of the shielding material. Therefore, the present invention does not necessarily require that the signal always penetrates the shielding material, and includes cases where a candidate person or animal can be estimated based on channel state information or equivalent propagation path information obtained through the shielding material.

[0026] The transmitting radio device 10 transmits a radio signal passing through the monitoring space 40. The receiving radio device 20 receives the radio signal and outputs RSSI and CSI. CSI may be the I / Q components, amplitude, phase of multiple subcarriers, or feature quantities obtained therefrom.

[0027] The local processing unit 50 analyzes the CSI received from the receiving wireless device 20 locally. The local processing unit 50 does not transmit the raw CSI data to an external cloud. Even when a notification is issued, the notification targets summary information such as risk level, status label, number of people estimate, animal candidates, and AI comments, not the I / Q column or subcarrier column itself. If the risk level is determined to be "dangerous," a notification can be sent externally via email or to a mobile phone via the notification unit 70.

[0028] As shown in Figure 2, the local processing unit 50 includes a CSI acquisition unit 101, a frame type filter 102, a CSI shaping unit 103, a vacant room criterion generation unit 104, a difference map generation unit 105, a motion analysis unit 106, a respiration estimation unit 107, a person estimation unit 108, an animal candidate estimation unit 109, a three-dimensional display control unit 110, a local AI comment unit 111, a notification control unit 112, and an external notification transmission unit 113. These can be implemented as software, firmware, a dedicated circuit, or a combination thereof.

[0029] The frame type filter 102 clusters the shape or statistical features of received frames and selects the majority frame type for analysis. This suppresses the overestimation of motion metrics due to the mixing of beacons, broadcasts, management frames, etc.

[0030] The CSI shaping unit 103 generates an amplitude column from the I / Q column, removes null subcarriers, suppresses outliers using a Hampel filter or the like, removes trends using a moving average or the like, and normalizes each subcarrier.

[0031] The vacancy criterion generation unit 104 generates vacancy criterion information using the criterion state immediately after startup or during a predetermined period. The difference map generation unit 105 calculates the difference between the current subcarrier information and the vacancy criterion information and generates a subcarrier deviation distribution.

[0032] The motion analysis unit 106 calculates a robust motion index using the correlation distance between consecutive packets or an equivalent difference index. The motion index is converted into short-term large movements, small movements, motion occupancy rates, higher quantile intensity, etc., and used for estimating the number of people and candidate animals.

[0033] The respiration estimation unit 107 resamples the CSI amplitude sequence within a predetermined time window at equal intervals, performs outlier removal and trend removal, and then calculates the spectrum of the respiration frequency band. The respiration estimation unit 107 uses a combination of peak extraction by FFT and periodic matching by the autocorrelation function (ACF) to calculate respiration quality and representative respiration peaks.

[0034] The representative respiratory peak is a representative value obtained from the reflective components mixed throughout the entire monitoring space and does not necessarily represent the individual respiratory rate of a specific person. Therefore, in this embodiment, the display control unit 110 explicitly indicates the representative respiratory peak as a "representative value" and displays a representative indicator above the head of the person candidate P1, etc., to indicate that it is an assigned display.

[0035] The person estimation unit 108 integrates multiple sustained respiratory candidate peaks, the occupancy rate of motion indicators, the upper quantile intensity, and the spread of the subcarrier deviation distribution to estimate the number of person candidates as a stepped value such as 0, 1, 2, or 3 or more. Hysteresis is provided during increases and decreases to suppress sudden spikes or drops in the display.

[0036] The animal candidate estimation unit 109 evaluates features corresponding to height, gait, or reflectors separately from the human candidates. For example, it scores the animal candidates by using the fact that the difference is local, the motion index shows short bursts of low to moderate magnitude, the spread of the subcarrier deviation is narrower compared to the human candidates, and there is a respiratory band or short-period variation.

[0037] The display of possible animals does not definitively confirm that it is a dog, but rather indicates to the user the possibility of an animal or animal reflector. However, in the case of monitoring dogs within the home, since the user already knows that the subject of monitoring is a dog, this display of possible animals is useful for identifying the presence or movement of a dog.

[0038] As shown in Figure 4, the three-dimensional display control unit 110 displays the transmitting wireless device 10, receiving wireless device 20, propagation path, Fresnel zone, voxel point cloud, pseudo-skeleton of human candidates, pseudo-quadrupedal display of animal candidates, representative indicator of representative respiratory peak, respiratory quality, RSSI, motion indicator, number of people estimate, animal candidate confidence level, etc., on the same local screen.

[0039] Three-dimensional display is not limited to survey results in actual coordinates. In the case of a single TX-RX link, CSI is a mixture of multiple reflection components in space, so the pseudo-skeleton and animal candidate displays are representations to visualize the candidate state in an easily understandable way. In embodiments using multiple links, the display position or candidate separation accuracy can be improved.

[0040] The local AI comment unit 111 takes inputs such as the number of people estimated, animal candidates, representative respiratory peak, respiratory quality, movement indicators, risk level, and notification control status, and generates natural language comments using a local inference model or rule-based processing. This makes it easier for users to understand the meaning of the situation, not just the numerical values.

[0041] The notification control unit 112 generates a summary notification when the level of danger exceeds a predetermined value or when a predetermined condition persists for a certain period of time. The notification control unit 112 suppresses excessive warning displays or notifications by using the notification prohibition period immediately after startup, the cool-down period, the time when breathing is not confirmed, the time when the device remains still after a large movement, etc.

[0042] The external notification transmission unit 113 sends a summary notification to an external recipient via email, SMS, MMS, push notification to a mobile phone terminal, or a notification service corresponding to a mobile phone number or email address, when the notification control unit 112 determines that the risk level is "dangerous". The summary notification includes at least one of the following: risk level, status label, time, number of people estimated, animal candidate, representative respiratory peak, respiratory quality, local AI comment, and recommended response, but does not include CSI raw data.

[0043] In the processing flow shown in Figure 3, the CSI is first acquired, the frame type is selected, and I / Q amplitude conversion and null subcarrier removal are performed. Next, the difference from the vacant room criterion, respiratory cycle, motion index, and deviation distribution are calculated, and the number of human candidates and animal candidates is estimated. Finally, a three-dimensional display, representative peak display, local AI comments, and summary notification are generated.

[0044] As shown in Figure 5, in respiration estimation, the representative respiratory peak and individual candidate peaks are treated separately. The representative respiratory peak is displayed on the screen as the confirmed respiratory count or representative respiratory count. On the other hand, individual candidate peaks can be displayed as candidate values ​​or as unseparated, such as P1 candidate, P2 candidate, P3 candidate, etc.

[0045] As shown in Figure 6, both the population estimation and the candidate animal estimation use features obtained from the same CSI, but the evaluation criteria differ. The population estimation emphasizes spread and multiple peaks, while the candidate animal estimation emphasizes differences in body height, gait, locality, duration, and reflectance distribution.

[0046] As shown in Figure 8, the present invention can be used in remote and communication-restricted environments such as home monitoring, animal monitoring, nursing homes, hospitals, vehicles, construction sites, disaster shelters, livestock barns, agricultural facilities, warehouses, exhibition facilities, space, or the lunar surface.

[0047] The present invention is not limited to a single TX-RX link. By using a network of multiple transmitting radio devices, multiple receiving radio devices, multiple frequency bands, multiple antennas, or multiple rooms, the separation accuracy of human candidates, animal candidates, breathing candidates, and location candidates can be improved.

[0048] This invention is applicable to dogs, cats, livestock, service animals, display animals, pets, and all other animals. Furthermore, when both human and animal candidates exist, they can be displayed as separate layers on the same screen.

[0049] The present invention is not limited to medical diagnostic devices. Respiratory rate and respiratory quality are estimated values ​​for monitoring assistance, and when intended for diagnostic purposes, the device can be configured to meet additional requirements as a medical device.

[0050] The present invention can use small language models, classifiers, rule-based AI, time-series models, neural networks, or combinations thereof as local inference models. The local inference model can generate comments that explain not only the level of risk but also why it is a cause for concern or normal.

[0051] The present invention may be independently understood in future divisional applications as inventions for estimating the number of human candidates, estimating animal candidates, displaying representative respiratory peaks, 3D pseudo-display, fully local AI commentary, external notification during risk assessment, notification control, applications to vehicles or construction sites, applications to space or the lunar surface, and improved separation accuracy through multiple links. [Examples]

[0052] An embodiment of the present invention will be described below. The fully local wireless sensing device of this embodiment comprises a transmitting wireless device 10, a receiving wireless device 20, a local processing unit 50, a display unit 60, and a notification unit 70. The transmitting wireless device 10 and the receiving wireless device 20 are composed of wireless communication modules such as ESP32-S3, and the local processing unit 50 is composed of a small computer such as a Raspberry Pi.

[0053] The local processing unit 50 generates vacant room reference information from a reference state after startup or during a predetermined period, and calculates the deviation from the current channel state information. It also calculates the correlation distance between consecutive packets, the subcarrier deviation distribution, the motion occupancy rate, and the periodic component of the respiratory frequency band. Based on this, it estimates the presence or absence of human candidates, the number of human candidates, the representative respiratory peak, and animal candidates.

[0054] In estimating the number of potential individuals, multiple peaks appearing in the respiratory band, motion occupancy, subcarrier deviation spread, and motion intensity are combined to perform a staged estimation of 0, 1, 2, or 3 or more individuals. A single transmit / receive link does not provide precise location or personal identification of individual persons; rather, it processes them as candidate displays to clearly indicate the state of the monitored space to the user.

[0055] In estimating animal candidates, we use reflective characteristics that differ from those of human candidates, characteristics corresponding to height, temporal variations in gait, local subcarrier deviations, and differences in short-term movement and reflective distribution. Animal candidates are not limited to dogs, but include cats, livestock, service animals, exhibition animals, pets, and other animals.

[0056] The display unit 60 displays the transmitting wireless device 10, the receiving wireless device 20, the monitoring space 40, a simulated skeleton of a candidate person, a simulated display of a candidate animal, a voxel point cloud, a representative respiratory peak, and the risk level in the same three-dimensional display space. The representative respiratory peak is distinguished from multiple candidate respiratory counts and displayed in a way that allows the user to understand that it is a representative value.

[0057] The local processing unit 50 generates natural language comments using a local inference model or rule-based processing based on the estimated number of people, candidate animals, representative respiratory peaks, respiratory quality, movement indicators, vacancy differences, and risk level. This comment generation is performed within the local processing unit 50 without going through an external cloud.

[0058] The notification unit 70 suppresses notifications when the risk level is normal or under observation, and sends a summary notification externally when the risk level is determined to be dangerous. The summary notification includes at least one of the following: risk level, status label, time, number of people estimated, animal candidates, representative respiratory peak, respiratory quality, local AI comment, and recommended response. The summary notification is sent via email, SMS, MMS, push notification to mobile phone terminals, or notification service corresponding to a mobile phone number or email address.

[0059] The notification unit 70 does not transmit the raw channel status information, I / Q column, or subcarrier column itself to the outside. This allows for the notification of dangerous conditions to the outside only when necessary, while protecting the privacy of the person or animal being monitored.

[0060] According to this embodiment, it is possible to present potential individuals, multiple potential individuals, representative respiratory peaks, and potential animals in the same three-dimensional display space without relying on cameras, wearable sensors, or external clouds. Furthermore, since external notifications are only issued when the risk level is high, it is possible to provide useful information for monitoring purposes while suppressing excessive notifications. [Industrial applicability]

[0061] This invention can be used in industrial fields to monitor people and animals without cameras in environments where external communication is restricted, such as homes, nursing homes, hospitals, accommodation facilities, construction sites, factories, warehouses, vehicles, ships, agricultural facilities, livestock barns, exhibition facilities, disaster shelters, space, or the lunar surface.

[0062] . This invention can be expanded in future divisional applications to individual industrial fields such as fully local wireless sensing, multiple person candidate estimation, animal candidate estimation, representative respiratory peak display, local AI comments, danger notifications via email or mobile phone, in-vehicle monitoring, construction site safety, disaster monitoring, and agricultural or livestock management. [Explanation of symbols]

[0063] 1. Fully Local Wireless Sensing Device 10. Transmitting radio equipment 20 Receiving radio equipment 30. Radio propagation paths 40 Surveillance space 50 Local Processing Units 60 Display 70 Notification Department 101 CSI Acquisition Department 102 Frame Type Filter 103 CSI Plastic Surgery Department 104 Vacancy Criteria Generation Unit 105 Difference Map Generation Unit 106 Motion Analysis Unit 107 Respiration estimation part 108 Number of people estimation section 109 Animal candidate estimation section 110 Three-dimensional display control unit 111 Local AI Comment Section 112 Notification Control Unit 113 External Notification Transmission Unit

[0064] As used herein, "Linux" is a registered trademark or trademark of Linus Torvalds in Japan and other countries. Additionally, "ESP32," "ESP32-S3," "Raspberry Pi," "Wi-Fi," "Bluetooth," "Python," "Ollama," "Gemma," and other company names, product names, service names, standard names, or software names may be trademarks or registered trademarks of their respective owners.

[0065] The present invention is not limited to the above-mentioned specific product names, service names, standard names, semiconductor module names, computer names, operating system names, programming language names, or inference model names, but can be implemented using equivalent wireless communication modules, wireless communication methods, local processing units, operating systems, program execution environments, inference models, or rule-based processing.

[0066] Even when specific product or service names are given as examples in this specification, they are merely examples to facilitate understanding of the present invention and are not intended to imply that the use of such products or services is an essential component.

Claims

1. An acquisition unit that acquires channel status information relating to radio signals transmitted from a transmitting radio device and received by a receiving radio device, A formatting unit that formats the channel state information as time-series subcarrier information, A difference calculation unit that calculates the difference between vacancy standard information based on the standard state at startup or during a predetermined period and the current subcarrier information, A respiratory estimation unit that extracts periodic components of the respiratory frequency band from the aforementioned subcarrier information and determines a representative respiratory peak, A living organism estimation unit estimates the number of human candidates and animal candidates based on the difference, the periodic component, the motion index of the channel state information, and the subcarrier deviation distribution. A display control unit generates a display of human candidates corresponding to the number of human candidates and a display of animal candidates corresponding to the number of animal candidates in the same three-dimensional display space, and displays the representative respiratory peak separately from the individual candidates. The acquisition unit, the shaping unit, the difference calculation unit, the respiration estimation unit, the living organism estimation unit, and the display control unit are operated locally without transmitting the raw data of the channel state information to an external cloud, A fully local wireless sensing device equipped with [specific features / features].

2. The fully local wireless sensing device according to claim 1, wherein the shaping unit clusters the frame shapes of the channel state information and selects the channel state information corresponding to the majority frame type as the target for analysis.

3. The shaping unit generates an amplitude sequence from the I component and Q component of the channel state information, removes null subcarriers, and performs outlier removal and trend removal, as described in claim 1.

4. The fully local wireless sensing device according to claim 1, wherein the respiration estimation unit resamples the subcarrier information at equal intervals and determines the representative respiration peak by using a combination of frequency spectral analysis and autocorrelation analysis in the respiration band.

5. The fully local wireless sensing device according to claim 1, wherein the living organism estimation unit estimates the number of candidate people as one, two, or three or more values ​​using the persistence of multiple candidate respiratory peaks, the occupancy rate of a motion index, the upper quantile intensity of the motion index, and the spread of the subcarrier deviation distribution.

6. The fully local wireless sensing device according to claim 1, wherein the living organism estimation unit estimates an animal candidate different from the human candidate display based on at least one of the difference, the motion index, the reflection characteristics corresponding to body height, the time variation characteristics corresponding to gait, or the subcarrier deviation characteristics corresponding to body surface scattering.

7. The fully local wireless sensing device according to claim 1, wherein the display control unit displays a transmitting wireless device, a receiving wireless device, a propagation path, a voxel point cloud, a pseudo-skeleton of a candidate person, a pseudo-quadrupedal display of a candidate animal, and a representative index indicating the assignment destination of the representative respiratory peak in the three-dimensional display space.

8. The fully local wireless sensing device according to claim 1, wherein the control unit takes at least one of the number of human candidates, the animal candidates, the representative respiratory peak, respiratory quality, movement index, and risk level as input and generates natural language comments using a local inference model or local rules.

9. The control unit further comprises a risk determination unit for determining the degree of risk and a notification control unit, The fully local wireless sensing device according to claim 1, wherein the notification control unit, when it determines that the risk level is dangerous, transmits a summary notification to an external party via email, SMS, MMS, push notification for a mobile phone terminal, or a notification service corresponding to a mobile phone number or email address, which includes at least one of the risk level, status label, time, number of person candidates, animal candidates, representative respiratory peak, respiratory quality, local AI comment, and response recommendation text, and which does not include the raw data of the channel status information.

10. A process of locally acquiring channel status information regarding a radio signal transmitted from a transmitting radio device and received by a receiving radio device, The process of formatting the channel state information as time-series subcarrier information, A process of calculating the difference between vacancy standard information based on the standard status and the current subcarrier information, A step of extracting periodic components of the respiratory frequency band from the aforementioned subcarrier information and determining a representative respiratory peak, A step of estimating the number of human candidates and animal candidates based on the difference, the periodic component, the motion index, and the subcarrier deviation distribution, The process involves generating a display of human candidates corresponding to the number of human candidates and a display of animal candidates corresponding to the number of animal candidates in the same three-dimensional display space, and displaying the representative respiratory peak separately from the individual candidates. A fully local wireless sensing method that includes the above, and is performed without transmitting the raw data of the channel status information to an external cloud.

11. A function that formats channel state information regarding radio signals between transmitting and receiving wireless devices as time-series subcarrier information. A function to calculate the difference between vacancy criteria information based on the standard status and the current subcarrier information. A function to extract periodic components of the respiratory frequency band from the aforementioned subcarrier information and determine a representative respiratory peak. A function to estimate the number of human candidates and animal candidates based on the aforementioned difference, the aforementioned periodic component, the motion index, and the subcarrier deviation distribution. Furthermore, a function to generate a display of human candidates corresponding to the number of human candidates and a display of animal candidates corresponding to the number of animal candidates in the same three-dimensional display space, and to display the representative respiratory peak separately from the individual candidates. A program that achieves this without sending the raw channel status information data to an external cloud.

Images