A human respiration detection method based on multi-subcarrier fusion

Abstract

The present application belongs to the field of radio technology, and relates to a human respiration detection method based on multi-subcarrier fusion, which comprises the following steps: a sending device is configured with at least one antenna, a receiving device is configured with three antennas, and channel state information (CSI) corresponding to one sending antenna in the receiving device is extracted; channel state information in radio frequency signals received by any two receiving antennas corresponding to the same sending antenna is divided to obtain new channel state information; after pre-processing and standardization processing, subcarriers with good sensing performance are selected for fusion through a power spectral density analysis method according to the new channel state information; phase calibration is performed on the subcarriers before fusion to obtain an optimal respiration detection signal; and a target respiration rate is obtained by calculating a power spectral density peak frequency of the optimal respiration detection signal. The human respiration detection method provided by the present application has good stability in the respiration detection process, and greatly expands the sensing range of respiration detection.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of radio, and particularly relates to a human respiration detection method based on multi-subcarrier fusion. BACKGROUND

[0002] With the rapid development of wireless communication technology, the technology of sensing detection using wireless signals has attracted widespread attention. Especially with the large-scale deployment of base stations, WiFi signals are ubiquitous in people's production and life, so the technology of integrating sensing and communication using WiFi signals has attracted much attention.

[0003] WiFi signal-based sensing is a non-contact sensing, which means that the target can be sensed without carrying any device or having any contact with the device. For example, in the process of human respiration detection, the respiration rate can be obtained without contacting the human body. Traditional contact-type and wearable respiration detection not only has high equipment cost and low anti-interference performance, but also needs to be carried, while non-contact sensing using WiFi has the advantages of convenience, low cost and non-intrusiveness, etc. Therefore, using WiFi signals for respiration detection can better meet the needs of daily health monitoring.

[0004] Early researchers used the received signal strength (RSS) of WiFi to sense the signal fluctuation caused by respiration, but due to the limitation of signal granularity, RSS is not sensitive to the signal change caused by human chest movement, and such signal change is easily overwhelmed by noise, so it is impossible to reliably detect respiration in practical applications. Since 2010, researchers have modified the firmware to enable the collection of channel state information (CSI) on ordinary Wi-Fi devices. Compared with RSS, CSI is a fine-grained information and has relatively stable overall structural features, which can sense more subtle and larger range of environmental information in time and frequency domains, so researchers have used CSI for respiration detection.

[0005] Current research on respiration detection based on WiFi channel state information mostly focuses on how to reduce noise in the signal and how to enhance the respiration reflection signal. However, the subcarriers used for respiration detection are random or single, and sometimes the subcarriers do not carry the respiration reflection signal or are completely covered by noise, resulting in low robustness of detection. Although some researchers have proposed a method of using multi-subcarrier fusion to solve the problem of single subcarrier, there is no much research on calibration before subcarrier fusion, resulting in that the fused signal still cannot reliably perform long-distance non-contact respiration detection. SUMMARY

[0006] In order to improve the above-mentioned existing respiratory detection method, the present application proposes a human respiratory detection method based on multi-subcarrier fusion, which can effectively detect the respiratory rate of human body at a long distance in a non-contact sensing mode.

[0007] In order to achieve the above-mentioned target, the technical scheme adopted by the present application is:

[0008] A human respiratory detection method based on multi-subcarrier fusion, the method comprises the following steps:

[0009] 1) The sending device sends a radio frequency signal to the receiving device; wherein the sending device is configured with at least one transmitting antenna, and the receiving device is configured with three receiving antennas; the radio frequency signal sent by the transmitting antenna carries a target respiratory signal which is received by the receiving antenna for target respiratory detection;

[0010] 2) The three receiving antennas of the receiving device contain the channel state information (CSI) of multiple subcarriers corresponding to each transmitting antenna; the channel state information (CSI) corresponding to the same transmitting antenna in the three receiving antennas is extracted; the channel state information (CSI) is in the form of 1x3x30, wherein 1 represents the number of transmitting antennas, 3 represents the number of receiving antennas, and 30 represents the number of subcarriers in the CSI corresponding to a pair of transmitting and receiving antennas;

[0011] 3) Divide the channel state information (CSI) in the radio frequency signals received by any two receiving antennas corresponding to the same transmitting antenna to obtain new channel state information; the new channel state information is in the same form as in step 2);

[0012] 4) Reduce the dimension of the complex matrix of the new channel state information and calculate the modulus of each complex number in the reduced matrix to extract the amplitude signal in the new channel state information;

[0013] 5) Preprocess the amplitude signal: filter out abnormal values in the amplitude signal and perform filtering and smoothing to obtain the preprocessed amplitude signal;

[0014] 6) Remove the baseline drift of the preprocessed amplitude signal; the method comprises: using Z-score method standardization processing to obtain the standardized amplitude signal, and then using median filter method to remove the baseline drift;

[0015] 7) Perform power spectral density analysis on the amplitude signal processed in step (6) to screen subcarriers with good sensing performance, including: calculating the power spectral density of the amplitude signal in a period of time, and taking the peak value PSD(n=1,2,…,90) of the power spectral density as the characteristic of the subcarrier n; the peak value of the power spectral density is the maximum value of the power spectral density; n ​

[0016] 8) the power spectral density peak value PSD of the amplitude signal obtained after step (7) is processed i Taking the maximum value, the power spectral density peak value PSD is selected n The subcarrier corresponding to the maximum value in (n = 1, 2, …, 90)

[0017]

[0018] wherein, is the subcarrier number corresponding to the maximum power spectral density peak value;

[0019] 9) According to the maximum value in the subcarrier power spectral density peak value, the fused subcarriers are selected from all subcarriers to form a set J;

[0020] 10) The phase calibration is performed on the set J, and the method comprises: respectively calculating the Euclidean distance between the subcarriers in the set J and the subcarrier corresponding to the maximum value in the power spectral density peak value The Euclidean distance is calculated, and according to the distance threshold, the subcarrier amplitude signal greater than the distance threshold is taken as the opposite to obtain a new set JX;

[0021] 11) The fusion is performed on all subcarriers in the set JX to obtain an optimal respiratory detection signal;

[0022] 12) The power spectral density is calculated for the optimal respiratory detection signal, and the frequency corresponding to the power spectral density peak value is the target respiratory rate, which is expressed in times per minute.

[0023] Further, the transmitting antenna and the receiving antenna are both vertically polarized omnidirectional antennas.

[0024] Further, the amplitude signal in step 4) contains 90 subcarriers, and the amplitude signal is mathematically expressed as {x(n)|n = 1, 2, …, 90}.

[0025] Further, step 5) specifically comprises:

[0026] 5-1) The Hampel filter is used to process the amplitude signal to filter out abnormal values in the amplitude signal;

[0027] 5-2) The S-G filtering method based on the least square method is used to filter and smooth the amplitude signal to obtain a preprocessed amplitude signal;

[0028] Further, the filter window size of the Hampel filter in step 5-1) is set to 50, and the threshold parameter is taken as 0.

[0029] Further, the filter window size of the S-G filter in step 5-2) is set to 201, and the fitting polynomial degree is set to 3.

[0030] Further, in step 7), the calculation method of the power spectral density is that the amplitude time-domain signal is converted into a frequency-domain signal by fast Fourier transform, and the power spectral density corresponding to the amplitude signal in a period of time is obtained.

[0031] Further, in step 9), the fused subcarriers are selected from all subcarriers according to the maximum value in the peak value of the power spectral density, to form a set J, which specifically includes the following operations: setting a threshold value of the peak value of the power spectral density, screening the subcarriers with the peak value of the power spectral density greater than the threshold value to form the set J; the threshold value is max(PSD n )-0.1.

[0032] Further, in step 10), the distance threshold value is 100, and the subcarrier amplitude signal greater than the distance threshold value is taken as -x(m), where m represents the serial number of the subcarrier greater than the distance threshold value;

[0033] Further, in step 11), all subcarriers in the set JX are fused to obtain an optimal respiratory detection signal, and the specific fusion method is that all subcarriers in the set JX are summed:

[0034]

[0035] wherein JX k represents the subcarrier k, and p is the number of subcarriers in the set JX.

[0036] Compared with the prior art, the present application has the following advantages:

[0037] The present application provides a human respiratory detection method based on multi-subcarrier fusion, which uses multiple pairs of antenna subcarriers, uses the power spectral density peak value to screen the subcarriers, eliminates the influence of subcarriers with poor sensing performance, and fuses and uses the screened subcarriers after phase calibration, which not only improves the stability of respiratory detection, but also greatly expands the sensing range of respiratory detection. In addition, in practical application, the technical scheme provided by the present application also has the advantages of simple method, low cost, anti-interference and the like. BRIEF DESCRIPTION OF DRAWINGS

[0038] Figure 1 is a flow block diagram of a human respiratory detection method based on multi-subcarrier fusion in an embodiment of the present application.

[0039] Figure 2 is a specific implementation schematic diagram of a human respiratory detection method in an embodiment of the present application.

[0040] Figure 3a This is a waveform diagram of the new CSI amplitude signal before preprocessing in an embodiment of the present invention.

[0041] Figure 3b This is a waveform diagram of the new CSI amplitude signal after preprocessing in an embodiment of the present invention.

[0042] Figure 4a This is a waveform diagram of the amplitude signal before value filtering in an embodiment of the present invention.

[0043] Figure 4b This is the signal power spectral density diagram of the amplitude signal before value filtering in an embodiment of the present invention.

[0044] Figure 4c This is a waveform diagram of the amplitude signal after value filtering in an embodiment of the present invention.

[0045] Figure 4d This is the signal power spectral density diagram of the amplitude signal after value filtering in an embodiment of the present invention.

[0046] Figure 5a This is a standardized amplitude diagram of each subcarrier phase in set J before phase calibration in this embodiment of the invention.

[0047] Figure 5b This is a standardized amplitude diagram of each subcarrier phase calibrated in set J in this embodiment of the invention.

[0048] Figure 6a This is a waveform diagram of subcarrier fusion before calibration in an embodiment of the present invention.

[0049] Figure 6b This is a waveform diagram of the fusion after subcarrier calibration in an embodiment of the present invention.

[0050] Figure 7 This is the power spectral density diagram of the optimal respiratory detection signal in this embodiment of the invention. Detailed Implementation

[0051] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings, so that those skilled in the art can fully understand the implementation of the present invention and better define the scope of protection of the present invention.

[0052] This invention provides a method for detecting human respiration based on multi-subcarrier fusion. This method does not require the subject to carry any equipment; it utilizes WiFi transceivers deployed in the environment to detect the breathing rate of a subject sitting or lying down at a distance. The specific implementation of this invention will be described below:

[0053] like Figure 1 The diagram shown is a flowchart of a human respiration detection method according to a specific embodiment of the present invention.

[0054] The method of the application comprises: a sending end is configured with at least one transmitting antenna, transmits radio frequency signals to a receiving end, the receiving end is configured with three receiving antennas, constructs new channel state information by using channel state information corresponding to each pair of transmitting and receiving antennas, pre-processes amplitude signals, normalizes data, removes signal baseline drift by using a median filter, screens subcarriers by using a power spectral density analysis method, calibrates phases of the subcarrier set and fuses them, and calculates a respiratory rate according to the fused optimal respiratory detection signal. The sending end and the receiving end are WiFi signal transceiving devices; the channel state information corresponding to each pair of transmitting and receiving antennas is channel state information corresponding to the same transmitting antenna. The following will explain the implementation process of the above content:

[0055] 1) The sending device is configured with at least one transmitting antenna, and transmits radio frequency signals to a receiving device; the receiving device is configured with three receiving antennas;

[0056] The sending device and the receiving device are WiFi signal transceiving devices, which are placed in an environment as shown in Figure 2 The transmitting antenna and the receiving antenna are vertically polarized omnidirectional antennas, the radio frequency signals transmitted by the transmitting antenna carry target respiratory signals and are received by the receiving antenna, which is used for target respiratory detection.

[0057] 2) The three receiving antennas of the receiving end contain channel state information (CSI) of multiple subcarriers of multiple transmitting antennas, and the channel state information (CSI) corresponding to the same transmitting antenna in the three receiving antennas is extracted, which is in the form of 1x3x30, wherein 1 represents the number of transmitting antennas, 3 represents the number of receiving antennas, and 30 represents the number of subcarriers in the CSI corresponding to one pair of transmitting and receiving antennas.

[0058] 3) Divide the channel state information corresponding to each pair of transmitting and receiving antennas to obtain new channel state information, which eliminates part of the amplitude noise; the new channel state information is in the same form as in step 2.

[0059] 4) Extract the amplitude signal in the new channel state information, the specific process is to reduce the dimension of the new CSI complex matrix and calculate the modulus of each complex number in the reduced matrix, and the extracted amplitude signal can be mathematically represented as {x(n)|n=1,2,…,90}.

[0060] 5) Pre-process the amplitude signal, the signal before and after pre-processing is as shown in Figures 3a-3b The specific process includes the following:

[0061] 5-1) Use a Hampel filter to process the amplitude signal, filter out abnormal values in the amplitude signal, set the filter window size to 50, and set the threshold parameter to 0.

[0062] 5-2) Using S-G filter method based on local polynomial least squares fitting to filter and smooth the amplitude signal, set the filter window size to 201, and set the polynomial fitting degree to 3.

[0063] 6) Remove the baseline drift of the preprocessed amplitude signal, the purpose is to remove the low frequency noise of the signal, so as to increase the power spectral density peak value, as shown in Figures 4a-4d ; The specific method includes: using Z-score method for standardization processing to obtain the standardized amplitude signal, and then using median filter method to remove the baseline drift.

[0064] 7) Perform power spectral density analysis on the amplitude signal to screen subcarriers with good sensing performance, the specific method comprising: calculating the power spectral density of the amplitude signal in a period of time, and taking the peak value PSD n (n = 1, 2, …, 90) of the power spectral density as the characteristic of the subcarrier n; The power spectral density calculation method is: converting the amplitude time domain signal into frequency domain signal by fast Fourier transform to obtain its corresponding power spectral density; The peak value of the power spectral density is the maximum value of the power spectral density.

[0065] 8) Take the maximum value of the power spectral density peak value PSD n of the amplitude signal, and select the subcarrier corresponding to the maximum value of the power spectral density peak value.

[0066]

[0067] Wherein, is the subcarrier number corresponding to the maximum power spectral density peak value.

[0068] 9) Select the fused subcarriers from all subcarriers according to the maximum value of the subcarrier power spectral density peak value to form set J;

[0069] The said selecting the fused subcarriers from all subcarriers to form set J, specifically includes: setting the threshold value of the power spectral density peak value, screening the subcarriers with power spectral density peak value greater than the threshold value to form set J; The threshold value is max(PSD n )-0.1.

[0070] 10) Phase calibration is performed on the set J, and the subcarriers before and after calibration are as shown in Figures 5a-5b The method includes: respectively adjusting the phase of the subcarriers in set J and the subcarrier The Euclidean distance is calculated, the subcarrier amplitude signals greater than the threshold are negated to obtain a new set JX according to the distance threshold, the distance threshold is 100, and the operation of negating the subcarrier amplitude signals greater than the threshold is -x(m), wherein m represents the subcarrier serial number greater than the distance threshold;

[0071] 11) The optimal respiratory detection signal is obtained by fusing all the subcarriers in the set JX, as shown in the following formula: Figures 6a-6b The signal comparison chart obtained by fusing the subcarriers before and after the phase calibration is shown in the following figure; the subcarrier fusion method is to sum all the subcarriers in the set JX:

[0072]

[0073] wherein JX k represents the subcarrier k, and p is the number of subcarriers in the set JX.

[0074] 12) The power spectral density of the optimal respiratory detection signal is calculated, as shown in the following formula: Figure 7 The frequency corresponding to the peak value of the power spectral density is the target respiratory rate, and the unit is times / minute.

[0075] In this embodiment, the data acquisition device is a Mini PC, one Mini PC is used as a signal sending device, and the other Mini PC is used as a signal receiving device. According to the detection accuracy and precision requirements of the system, at least one transmitting antenna is configured on the sending end, and three receiving antennas are configured on the receiving end; it should be noted that, in order to achieve higher detection precision and stability, the selected channel state information can be increased to multiple transmitting antennas, so that the subcarriers with good sensing performance are selected from more subcarriers for fusion. The specific implementation site is a laboratory, and the LoS of the transceiver equipment is set to 1m-5m. When a person is sitting and lying in the breathing scene, the respiratory frequency of the person can be accurately detected.

[0076] Taking the transceiver LoS=4m as an example, in the laboratory, when a person is sitting or lying, the respiratory frequency of the person can be accurately detected, as shown in the following figure: Figure 2As shown, the data packet containing channel state information is collected for 60s each time at different breathing rates; a.dat file is obtained after collection; the CSI complex matrix is extracted from the.dat file, and the CSI amplitude signal is obtained after dimension reduction and modulus taking; then, 90 subcarriers in the CSI amplitude signal are preprocessed, and abnormal values and filtering smoothing are removed; subsequently, the Z-score method is used for standardization processing on the 90 subcarriers, and median filtering is performed on the standardized subcarriers to remove signal baseline drift; then, power spectral density analysis is performed on the 90 subcarriers; according to the power spectral density peak value threshold, a good-performing subcarrier set J is screened; the subcarriers in the set J are phase calibrated and fused to obtain an optimal respiratory detection signal; the power spectral density of the optimal respiratory detection signal is calculated to obtain a target breathing rate. Experimental results show that the present application can effectively detect human respiration at a long distance in a laboratory environment, a classroom and a corridor.

[0077] The human respiratory detection method provided by the present application not only has good stability in the respiratory detection process, but also greatly expands the sensing range of respiratory detection, and has the advantages of simple method, low cost and anti-interference in actual application.

[0078] The above is the specific implementation process of the present application, which is intended to facilitate the understanding of those skilled in the art and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation according to the content of the specification and drawings within the technical scope disclosed by the present application is included in the patent protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for human respiration detection based on multi-subcarrier fusion, characterized in that, The method comprises the following steps: 1) a sending device sends a radio frequency signal to a receiving device; wherein the sending device is configured with at least one transmitting antenna, and the receiving device is configured with three receiving antennas; the radio frequency signal sent by the transmitting antenna is received by the receiving antenna to carry a target respiration signal for target respiration detection; 2) the three receiving antennas of the receiving device comprise channel state information (CSI) of a plurality of subcarriers corresponding to each transmitting antenna; the channel state information (CSI) corresponding to the same transmitting antenna in the three receiving antennas is extracted; 3) the channel state information (CSI) in the radio frequency signals received by any two receiving antennas corresponding to the same transmitting antenna is divided to obtain new channel state information; 4) an amplitude signal in the new channel state information is extracted; specifically, dimensionality reduction is performed on the new CSI complex matrix, and the modulus of each complex number in the reduced matrix is calculated, and the extracted amplitude signal is mathematically represented as {x(n)|n=1,2,…,90}; 5) the amplitude signal is preprocessed: abnormal values in the amplitude signal are filtered out, and filtering and smoothing are performed to obtain a preprocessed amplitude signal; 6) the preprocessed amplitude signal obtained in step (5) is removed from the baseline drift; 7) performing power spectral density analysis on the amplitude signal processed in step (6) to screen subcarriers with good sensing performance, including: calculating the power spectral density of the amplitude signal in a period of time, and taking the peak value PSD of the power spectral density n (n=1,2,…,90) as the characteristic of subcarrier n; 8) the power spectral density peak PSD of the amplitude signal obtained after processing in step (7) n Taking the maximum value, the power spectral density peak PSD is selected n The subcarrier x(n = 1, 2, …, 90) corresponding to the maximum value in the sequence ): wherein n is a subcarrier index, is the subcarrier index corresponding to the maximum power spectral density peak. 9) according to the maximum value in the subcarrier power spectral density peak value, the fused subcarriers are selected from all subcarriers to form a set J; 10) Perform phase calibration on the set J, the method comprising: calibrating the subcarriers in set J with the subcarrier x(x) corresponding to the maximum value among the peak power spectral density values. ) Calculate the Euclidean distance, and based on the distance threshold, invert the subcarrier amplitude signals that are greater than the distance threshold to obtain a new set JX; 11) all subcarriers in the set JX are fused to obtain an optimal respiration detection signal; 12) the power spectral density of the optimal respiration detection signal is calculated, and the frequency corresponding to the power spectral density peak value is the target respiration rate, with units of times / minute; In step 9), the fused subcarriers are selected from all the subcarriers according to the maximum value in the peak value of the power spectrum density of the subcarriers to form a set J, which specifically includes the following operations: setting a threshold value of the peak value of the power spectrum density, screening the subcarriers with the peak value of the power spectrum density greater than the threshold value to form the set J; the threshold value is ; In step 10), the distance threshold value is in the range of 80-100; In step 11), the all subcarriers in the set JX are fused to obtain an optimal respiration detection signal, and the specific fusion method is to sum all subcarriers in the set JX: wherein denotes a subcarrier k, p is the number of subcarriers in the set Jx.

2. The multi-subcarrier fusion based human respiration detection method of claim 1, wherein: Both the transmitting antenna and the receiving antenna are vertically polarized omnidirectional antennas.

3. The multi-subcarrier fusion based human respiration detection method of claim 1, wherein, Step 5) specifically includes: 5-1) the Hampel filter is used to process the amplitude signal to filter out abnormal values in the amplitude signal; 5-2) the S-G filtering method based on the least square method is used to filter and smooth the amplitude signal to obtain a preprocessed amplitude signal.

4. The method of claim 3, wherein the plurality of subcarriers are fused. In step 5-1), the filter window size of the Hampel filter can be set in the range of 50-80, and the threshold parameter value is 0.

5. The method of claim 3, wherein the plurality of subcarriers are fused to detect the respiration of the human body. In step 5-2), the filter window size of the S-G filter is set in the range of 201-301, and the fitting polynomial degree is set to 3.

6. The method of claim 1, wherein the plurality of subcarriers are fused. In step 7), the calculation method of the power spectral density is to convert the amplitude time domain signal into a frequency domain signal through fast Fourier transform to obtain the power spectral density corresponding to the amplitude signal in a period of time.

Images