A method for detecting the boundary of human respiration perception based on channel state information

Abstract

A method for detecting the boundary of human breathing perception based on channel state information, and the specific implementation steps include: First, collect the wireless signal data corresponding to human breathing in an indoor environment, and extract the original CSI data from the WiFi signal; Second, perform data preprocessing on the original CSI data, and the preprocessing steps include: outlier removal and noise filtering; Then, for the preprocessed data, use subcarrier selection and Fast Fourier Transform (FFT) to obtain the breathing frequency; Next, model the processed data to derive the conditions for the theoretically detected boundary of the breathing frequency; Finally, conduct tests based on the model to achieve the detection of the boundary of the human breathing frequency. This method realizes the perception of human breathing behavior and the estimation of the breathing rate by analyzing and processing the channel state information CSI. Combining with the breathing detection boundary model, it realizes the detection of the boundary of the human breathing frequency.

Description

Technical Field

[0001] The present invention relates to the technical field of human perception and respiratory boundary detection, and particularly relates to a method for detecting the respiratory perception boundary of a person based on channel state information. Background Art

[0002] In recent years, the vital sign monitoring system based on radio frequency signals has attracted much attention due to its advantages such as non-invasiveness and privacy protection. The principle behind these systems is that the minute movement of the chest during breathing causes minute changes in the propagation path of the nearby radio signals. Some systems utilize dedicated radio frequency (RF) devices, such as universal software radio peripherals, frequency modulated carrier radars, Doppler radars, and ultra-wideband multiple input multiple output (MIMO) radars. Although the solutions based on these dedicated devices have proven their effectiveness and high accuracy, due to the relatively high deployment cost and the dependence on special hardware, the application of these solutions in daily households has been severely hindered.

[0003] In order to meet the requirements of non-invasiveness and high cost performance, the inexpensive commercial WiFi devices widely distributed in daily life have recently been used by many researchers for non-contact respiratory perception. The early research solutions mainly used the received signal strength indicator (RSSI) of the WiFi signal to perceive the environment. The personnel respiratory perception solution based on RSSI usually requires the WiFi device to be close to the chest of the detector for respiratory perception, and can only provide coarse-grained perception. Recent research has shown that the channel state information (CSI) of the WiFi signal can be extracted from the open-source driver of the commercial WiFi chipset. Compared with the solution based on the RSSI signal, CSI provides more fine-grained physical layer subcarrier information and has the ability to remotely monitor the respiration of a person. Recently, a number of accurate personnel respiratory monitoring works based on CSI have been designed.

[0004] Although current CSI-based respiratory monitoring systems can already obtain accurate results, these efforts lack in-depth analysis and evaluation of the respiratory rate detection range. In other words, most existing work ignores the issue of evaluating the maximum distance at which a receiver can detect human respiration under conditions where respiration can be detected by a WiFi device. The study of the detection boundary of human respiratory rate is of great significance for non-contact vital sign perception, and its importance is mainly reflected in the following aspects. Second, it improves the convenience and usability of the system. Since respiratory monitoring is usually carried out continuously throughout the night and may last for several months or years, especially for those with chronic respiratory diseases, if the WiFi device is placed too close to the human body, it is very likely to affect the daily life of the subject. Summary of the Invention

[0005] Based on the above background technology, the present invention proposes a method for detecting the respiratory perception boundary of a person based on channel state information to reduce the interference of the device to people and improve the convenience of the device at the same time. In this method, after preprocessing the data, subcarrier selection and FFT are used to obtain the respiratory rate, and based on the modeling, the conditions for the theoretically detected respiratory rate boundary are derived and tested to achieve the detection of the respiratory rate boundary of a person.

[0006] A method for detecting the respiratory perception boundary of a person based on channel state information includes the following specific steps:

[0007] Step 1: Collect wireless signal data in the indoor environment under the respiratory state of a person, and extract the original channel state information CSI data therefrom;

[0008] Step 2: Perform data preprocessing on the CSI original data, including outlier removal and noise filtering;

[0009] Step 3: For the preprocessed data, use subcarrier selection and fast Fourier transform FFT to obtain the respiratory rate of the person;

[0010] Step 4: Model the processed data, establish an association model between the reflection path length of the wireless signal and the respiratory rate, estimate the amplitude attenuation and transmission delay of the reflection path, separate the respiratory component and the noise component in the CSI amplitude, and derive the respiratory rate detection boundary conditions;

[0011] Step 5: Perform tests based on the model to achieve the detection of the respiratory rate boundary of a person.

[0012] Furthermore, in Step 1, a communication device including a transmitter and a receiver is used to collect wireless signal data. The receiver and the transmitter are placed on both sides of the person, and both use the AP mode for wireless communication. The working frequency of WiFi is 2.4 GHz, and the receiver extracts the original channel state information CSI data from the collected WiFi signal.

[0013] Further, in step 2, the data preprocessing includes outlier removal and noise filtering, specifically as follows:

[0014] Step 2-1: Outlier removal; the Hampel identifier is used to remove outliers from the original data;

[0015] Step 2-2: Noise filtering; a band-pass filter is used to filter out the noise signal, and the frequency band of the received signal is limited to the normal breathing frequency band.

[0016] Further, in step 3, subcarrier selection and fast Fourier transform (FFT) are performed to obtain the human breathing frequency, specifically as follows:

[0017] Step 3-1: Subcarrier selection method; the variance of the CSI amplitude is calculated through a time window, and then the subcarrier with the largest variance is selected;

[0018] Step 3-2: Breathing frequency estimation; the fast Fourier transform (FFT) is performed on the subcarrier signal selected above to extract the breathing frequency of the experimenter.

[0019] Further, in step 4, the transmitter Tx and the receiver Rx are defined. The transmitter is placed on one side of the person's body, and the receiver is placed on the other side. Point A is the highest point of the abdomen corresponding to the person's maximum inhalation, and point B is the lowest point of the abdomen corresponding to the person's maximum exhalation; point P is any point between A and B, representing the abdominal position at any moment during the entire breathing process. The transmitted signal is reflected from point P to the receiver, and point C is the projection of point B on the horizontal plane where the transmitter is located. Point D is the projection of point C on the horizontal plane where the receiver is located; the distances between AB, BC, and CD are denoted as d AB , d BC , d CD ; point E is the intersection of the horizontal line of the receiver Rx and its perpendicular line passing through point D, and point F is the foot of the perpendicular from the line passing through the transmitter to the cross-section where the person's abdomen is located. d ER and d TF are the distances from the receiver and the transmitter to the cross-section where the experimenter's abdomen is located, respectively; the distance between FC is denoted as d FC .

[0020] Further, when d ER is 0, the receiver is directly opposite the experimenter's abdomen. When d ER is not 0, the receiver is not directly opposite the abdomen.

[0021] Further, when d CD is 0, the receiver and the experimenter are at the same height.

[0022] Further, in step 4, it specifically includes the following steps:

[0023] Step 4-1: Establish an association model between the reflection path length d TPR and the breathing frequency f BR ; The reflection path consists of two paths from the transmitter to the person's abdomen and from the person's abdomen to the receiver, that is, d TPR The expression is composed of d TP and d PR Two parts, and their Taylor series expansions are used to obtain the values of d TP and d PR , which are denoted as:

[0024]

[0025]

[0026] Step 4-2: Estimate the reflection path amplitude attenuation α TPR and the transmission delay τ TPR ; The reflection path length is obtained from step 4-1, and through Taylor formula expansion, the reflection path amplitude attenuation α TPR and the transmission delay τ TPR are obtained as follows:

[0027]

[0028] Where A TPR is a constant, G t , G r are the antenna gains of the transmitter and the receiver respectively, t is time, λ and f are the propagation wavelength and frequency of the radio signal in the dormitory scenario;

[0029] Step 4-3: Separate the breathing component and the noise component in the CSI amplitude; The multipath signal propagation path is expressed as the superposition of the direct path and the reflection path, and the channel impulse response CIR of the indoor wireless multipath channel is expressed as:

[0030]

[0031] Where α m (t), τ m (t) and θ m (t) are the amplitude attenuation, propagation delay and phase shift of the signal in the m-th path respectively, m is the TR and TPR, that is, the direct path and the reflection path; z(t) is the noise;

[0032] By performing a Fourier transform on h(t) over a period of time, the channel frequency response CFR for the corresponding time period is obtained; the H(f) over a period of time is measured, that is, the channel state information CSI. When performing a Fourier transform on a continuous CSI packet over a period of time, it is considered that α m (t), τ m (t) and θ m (t) are approximately invariant and are constants; the CSI measured by a single packet is expressed as:

[0033]

[0034] To facilitate the calculation of the CSI amplitude of the received signal, H kΔT (f) is expressed in complex form and simplified. The squared value of the CFR amplitude of the received signal obtained is denoted as:

[0035] |H kΔT (f)| 2 =α TR 2 +α TPR 2 +|Z kΔT (f)| 2 +2α TR α TPR cos(2πfτ TPR +θ TPR -2πfτ TR )+2α TR Z R (f)cos 2πfτ TR +2α TPR Z R (f)cos(2πfτ TPR +θ TPR )-2α TR Z I (f)sin 2πfτ TR -2α TPR Z I (f)sin(2πfτ TPR +θ TPR )

[0036] where Z R (f) and Z I (f) are the real and imaginary parts of the noise Z(f) respectively, j is the imaginary part symbol, and |·| represents the modulus of a complex number; the above formula is expanded by Taylor series and band-pass filtered to retain the signal within the breathing frequency band; then the DC component is removed to obtain the breathing frequency and the noise frequency; then a fast Fourier transform is performed to obtain 4 frequency points, namely the breathing signal frequency f BR , the noise signal frequency f, the mixed frequency of the breathing and noise signals f + fBR and f-f BR Set the amplitude corresponding to each frequency point to AmpZ f ,

[0037] Step 4-4: Solve the boundary d for detecting the breathing frequency of the person RFDB value; During the process of the transmitter gradually moving away from near the person's abdomen, the distance d between the transmitter and the person's abdomen TF gradually increases, resulting in the gradual increase of the reflection path length d TPR gradually increases, and the signal strength of the corresponding breathing signal will also gradually attenuate, specifically including the following 3 stages:

[0038] Stage 1, when the transmitter is close to the person's abdomen position, at this time the breathing signal is stronger than the noise signal, and it is considered that the transmitter is within the breathing frequency detection boundary at this time;

[0039] Stage 2, when the transmitter is far from the person's abdomen position, at this time the breathing signal will have a large attenuation, and the breathing signal strength will be equal to the noise signal strength, and it is considered that the transmitter is on the breathing frequency detection boundary at this time;

[0040] Stage 3, when the transmitter is very far from the person's abdomen position, the breathing signal attenuates severely, and the noise signal is stronger than the breathing signal, and it is considered that the transmitter is outside the breathing frequency detection boundary at this time;

[0041] Based on the above analysis, it is expressed as:

[0042]

[0043] where AmpZ max is the maximum value in AmpZ f , ;

[0044] When within the RFDB, although the breathing signal will attenuate with the increase of d TF , it is still stronger than the noise signal. Therefore, the amplitude of the breathing signal is the maximum value among the amplitudes corresponding to the 4 frequency points. And there are 3 frequency points corresponding to the noise. Only when the amplitude AmpZ max of the maximum frequency point corresponding to the noise signal is almost equal to the amplitude of the breathing signal , it is considered that the transmitter is on the RFDB. From this, it can be known that the RFDB condition is When this condition is satisfied, the d TF value obtained is the d RFDB value, that is

[0045]

[0046] By modeling and deriving the breathing frequency detection boundary, the conditions for determining whether the sensing device is at the breathing frequency detection boundary are obtained, and the theoretical breathing frequency detection boundary d of the person is solved. RFDB value.

[0047] Furthermore, in step 5, based on the model proposed in step 4, testing is carried out to implement the breathing frequency detection boundary, specifically:

[0048] Observe the spectrogram of the data collected by the receiver when the transmitter is at different positions; when the position of the transmitter is exactly on the detection boundary, a highest peak and a comparable sub-peak will appear in the spectrogram at this position; the determination condition for the breathing frequency detection boundary is given: when the intensity of the breathing signal is compared with the intensity of the noise signal, the sensing device is at the breathing frequency detection boundary;

[0049] The comparison between the intensity of the human breathing signal and the intensity of the noise signal needs to satisfy that the frequency difference corresponding to the highest peak and the sub-peak in the spectrogram is greater than 0.1 Hz, and the ratio of the difference obtained by subtracting the amplitude of the sub-peak from the amplitude of the highest peak to the amplitude of the sub-peak is less than 0.1; assuming that the frequencies of the two signals are f1 and f2, and the corresponding amplitudes are Amp1 and Amp2, if the two signals satisfy:

[0050] |f1 - f2| > 0.1 Hz, and

[0051] then it is considered that the two signals are compared with each other; when the highest peak and the sub-peak in the spectrogram are just compared, this position is considered as the detection boundary.

[0052] The beneficial effects of the present invention are:

[0053] (1) The present invention uses the channel state information CSI signal of WiFi, which can reflect the multipath propagation effect of wireless signals and perform fine-grained perception of the environment. Compared with traditional perception methods, the CSI signal perception method based on WiFi can achieve high-precision, peripheral-free, and user-contactless behavior perception, greatly improving the user experience.

[0054] (2) The present invention accurately describes and depicts the breathing detection boundary from the perspective of wireless signal multipath propagation, providing a theoretical basis for studying the detection boundaries of various fine-grained perception tasks and their system design.

[0055] (3) By analyzing the influence of human breathing on the wireless propagation channel, the present invention first designs a multipath channel model for breathing interference, and then further obtains a breathing frequency detection boundary model. The designed channel model reveals the perturbation characteristics of breathing frequency on the CSI amplitude, while the breathing frequency detection boundary model reveals the internal relationship between the transceiver position and the detectability of human breathing. Description of the Drawings

[0056] Figure 1 It is the flowchart of the breathing frequency detection boundary estimation system in the embodiment of the present invention.

[0057] Figure 2 It is the core algorithm diagram of the breathing frequency detection boundary estimation in the embodiment of the present invention.

[0058] Figure 3 It is the schematic diagram of the breathing frequency detection boundary estimation scenario in the embodiment of the present invention. Specific embodiments

[0059] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings of the specification.

[0060] A method for detecting the breathing frequency boundary of a person based on channel state information. First, collect the wireless signal data corresponding to a person's breathing in an indoor environment, and extract the original CSI data from the WiFi signal; secondly, perform data preprocessing on the original CSI data, and the preprocessing steps include: outlier removal and noise filtering; then, for the preprocessed data, use subcarrier selection and fast Fourier transform to obtain the breathing frequency; then, model the processed data to derive the conditions for the theoretically detected breathing frequency boundary; finally, perform tests on the basis of the model to achieve the detection of the breathing frequency boundary of a person. The present invention realizes the perception of a person's breathing behavior and the estimation of the breathing rate by analyzing and processing the channel state information CSI. Combining with the breathing detection boundary model, the detection of the breathing frequency boundary of a person is realized.

[0061] A method for detecting the breathing perception boundary based on channel state information, as follows Figure 1 , specifically including the following steps:

[0062] Step 1: Collect the wireless signal data in the state of a person's breathing in an indoor environment, and extract the original CSI data from it. The specific steps are as follows:

[0063] The communication devices used in this embodiment include a smart phone and a laptop computer equipped with an Intel 5300 network card. This computer is equipped with 3 external antennas. Among them, the smart phone is the transmitter and the laptop computer is the receiver. In order to better perceive the breathing behavior of the test subject, the receiver and the transmitter are placed on both sides of the test subject. The two devices perform wireless communication in AP mode, and the working frequency of WiFi is 2.4 GHz. The receiver extracts the original channel state information CSI data from the collected WiFi signal. The channel state information CSI represents the link change state of the wireless signal in space propagation and can reflect the highly sensitive changes in the surrounding environment.

[0064] Step 2: Perform data preprocessing on the original CSI data, including outlier removal and noise filtering. The specific steps are as follows:

[0065] Step 2-1: Outlier removal. Since the original data contains outlier data, which will affect the final estimation result. Therefore, this method first uses the Hampel identifier to remove outliers from the original data.

[0066] Step 2-2: Noise filtering. Use a band-pass filter to filter out the noise signal and limit the frequency band of the received signal within the normal breathing frequency band. Usually, the normal breathing frequency is between 0.2 Hz and 0.5 Hz.

[0067] Step 3: Perform subcarrier selection and fast Fourier transform (FFT) on the preprocessed data to obtain the breathing frequency of the person. The specific steps are as follows:

[0068] Step 3-1: Subcarrier selection method. Since the center frequencies of different subcarriers are different and their propagation paths in the air are also different, the sensing granularity of different subcarriers will also be different. Therefore, an effective subcarrier selection method is needed to screen out the subcarriers most suitable for breathing sensing. The specific subcarrier selection method is as follows: mainly through the variance of the signal amplitude var in a time window, which is expressed as:

[0069]

[0070] where N is the window size of the sliding window, represents the average value of the signal amplitudes in a sliding window, and |Amp(f) i | represents the i-th amplitude in the sliding window. Calculate the variance of the CSI amplitude through the above formula, and then select the subcarrier with the largest variance.

[0071] Step 3-2: Breathing frequency estimation. Perform a fast Fourier transform (FFT) on the selected subcarrier signal above. The fast Fourier transform can be calculated through the fft() function in MATLAB, and then extract the breathing frequency of the experimenter.

[0072] Step 4: Model the processed data and derive the theoretically breathing frequency detection boundary conditions. The specific steps are as follows:

[0073] Figure 2 is the core algorithm diagram for breathing frequency detection boundary estimation in the embodiments of the present invention.

[0074] The Respiration Frequency Detection Boundary (RFDB) is defined as the maximum distance between the transmitter and the abdomen of the person being detected, provided that the receiver can detect the respiration frequency of the person. It is usually denoted as d. RFDB 。

[0075] Figure 3 This is a scenario diagram of the method for detecting the respiration frequency boundary of a person in an embodiment of the present invention.

[0076] In Figure 3 ,the mobile phone serves as the transmitter Tx, and the external antenna of the laptop serves as the receiver Rx. The transmitter is placed on the bed on one side of the experimenter's body, and the receiver is placed on the table diagonally across the bed. Point A is the highest point of the abdomen corresponding to the experimenter's maximum inhalation, and point B is the lowest point of the abdomen corresponding to the experimenter's maximum exhalation. Point P is any point between A and B, representing the abdominal position at any moment during the entire breathing process. The transmitted signal can be reflected from point P to the receiver. Point C is the projection point of point B on the bed board. Point D is the projection point of point C on the table directly below the bed. The distances between AB, BC, and CD are denoted as d AB ,d BC ,d CD 。Point E is the intersection point of the horizontal line of the receiver Rx and the perpendicular line passing through point D. Point F is the foot of the perpendicular from the line passing through the transmitter to the cross-section where the experimenter's abdomen is located. d ER and d TF are the distances from the receiver and the transmitter to the cross-section where the experimenter's abdomen is located, respectively. The distance between FC is denoted as d FC 。

[0077] In Figure 3 ,the minimum value of d ER can be 0. When d ER is 0, the receiver is directly opposite the experimenter's abdomen. When d ER is not 0, the receiver is not directly opposite the abdomen. The minimum value of d CD can also be 0. When d CD is 0, the receiver and the experimenter are at the same height. In daily life, the receiver is generally not at the same height as the person, so the value of d CD is not 0.

[0078] Step 4-1: Establish an association model between the reflection path length d TPR and the respiration frequency f BR 。

[0079] First, the expression of d TPR is composed of d TP and d PRIt consists of two parts. By performing Taylor series expansion on them, the values of d TP and d PR are obtained and denoted as:

[0080]

[0081] Step 4-2: Estimate the reflection path amplitude attenuation α TPR and the transmission delay τ TPR .

[0082] From Step 4-1, the reflection path length can be obtained. Since the change in the reflection path caused by abdominal undulation in the indoor environment is a reflection process, it can be considered that the shadow fading is approximately constant and remains as a constant. By expanding through Taylor's formula, the reflection path amplitude attenuation α TPR and the transmission delay τ TPR are obtained. Their calculation results are as follows:

[0083]

[0084] where A TPR is a constant, G t , G r are the antenna gains of the transmitter and the receiver respectively, t is time, and λ and f are the propagation wavelength and frequency of the radio signal in the dormitory scenario.

[0085] Step 4-3: Separate the breathing component and the noise component in the CSI amplitude.

[0086] In the indoor environment, when the wireless signal transmitted by the transmitter encounters the human body, it will be propagated to the receiver in the form of reflection. Due to the different propagation paths of the multipath signals, the signal paths arriving at the receiver are superimposed on each other. For the convenience of representation, it is simplified as the superposition of the direct path and the reflection path. Therefore, the channel impulse response (CIR) of the indoor wireless multipath channel is expressed as:

[0087]

[0088] where α m (t), τ m (t) and θ m (t) are the amplitude attenuation, propagation delay and phase shift of the signal on the mth path (such as the TR and TPR paths) respectively. z(t) is the noise, and the two terms of the fractional expression are the direct path CIR and the reflection path CIR respectively.

[0089] By performing a Fourier transform on h(t) over a period of time, the channel frequency response (CFR) over the corresponding time period can be obtained. In practical applications, with the help of the CSI-tool software on the WiFi platform, H(f) over a period of time can be measured, which is the channel state information CSI. This period of time is usually only a few tens of microseconds because the duration of each CSI packet is very short. When performing a Fourier transform on consecutive CSI packets within an extremely short period of time, it can be considered that α m (t), τ m (t) and θ m (t) are approximately invariant and are constants. The CSI measured from a single packet can be expressed as:

[0090]

[0091] To facilitate the calculation of the CSI amplitude of the received signal, H kΔT (f) is expressed in complex form and simplified. The square of the CFR amplitude of the received signal obtained is denoted as:

[0092] |H kΔT (f)| 2 =α TR 2 +α TPR 2 +|Z kΔT (f)| 2 +2α TR α TPR cos(2πfτ TPR +θ TPR -2πfτ TR )+2α TR Z R (f)cos 2πfτ TR +2α TPR Z R (f)cos(2πfτ TPR +θ TPR )-2α TR Z I (f)sin 2πfτ TR -2α TPR Z I (f)sin(2πfτ TPR +θ TPR )

[0093] where Z R (f) and Z I(f) are the real and imaginary parts of the noise Z(f), j is the imaginary unit, and |·| represents the modulus of a complex number. Expand the above equation into a Taylor series and perform band-pass filtering to retain the signal within the breathing frequency band. Then remove the DC component to obtain the breathing frequency and the noise frequency. Then perform a fast Fourier transform to obtain 4 frequency points, which are the breathing signal frequency f BR , the noise signal frequency f, the mixing frequency of the breathing and noise signals f + f BR and f - f BR . Set the amplitude corresponding to each frequency point to AmpZf,

[0094] Step 4-4: Solve the boundary d RFDB for detecting the breathing frequency of the person.

[0095] During the process of the transmitter gradually moving away from near the abdomen of the experimenter, the distance d TF between the transmitter and the abdomen of the experimenter gradually increases, resulting in a gradual increase in the reflection path length d TPR , and the signal strength of the corresponding breathing signal will also gradually decay. Specifically, there are the following 3 stages:

[0096] (1) When the transmitter is close to the abdomen of the experimenter, the breathing signal is stronger than the noise signal at this time, and it can be considered that the transmitter is within the breathing frequency detection boundary at this time.

[0097] (2) When the transmitter is far from the abdomen of the experimenter, the breathing signal will have a large attenuation at this time, and the breathing signal strength will be equal to the noise signal strength. It can be considered that the transmitter is on the breathing frequency detection boundary at this time.

[0098] (3) When the transmitter is very far from the abdomen of the experimenter, the breathing signal decays severely, and the noise signal is very strong relative to the breathing signal. It can be considered that the transmitter is outside the breathing frequency detection boundary at this time.

[0099] Based on the above analysis, it can be expressed as:

[0100]

[0101] where AmpZ max is the maximum value in AmpZ f . When within the RFDB, although the breathing signal will decay as d

[0102] increases, it is still stronger than the noise signal. Therefore, the amplitude of the breathing signal TF ​is the maximum value among the amplitudes corresponding to 4 frequency points. There are 3 frequency points corresponding to the noise. Only when the amplitude AmpZ max of the noise signal is almost equal to the amplitude of the respiration signal can it be considered that the transmitter is on the RFDB. From this, it can be known that the RFDB condition is TF When this condition is met, the obtained d RFDB value is the d

[0103]

[0104] By modeling and theoretically deriving the boundary of respiration frequency detection, the condition for judging whether the sensing device is on the boundary of respiration frequency detection is obtained, and the theoretically obtained boundary d RFDB value of the respiration frequency detection of personnel is solved.

[0105] Step 5: Conduct tests based on the model to implement the respiration frequency detection boundary. The specific steps are as follows:

[0106] By observing the spectrogram of the data collected by the receiver when the transmitter is in different positions. When the position of the transmitter is exactly on the detection boundary, theoretically, there will be two highest peaks in the spectrogram at this position, which are the frequency peaks corresponding to the respiration signal and the noise signal respectively. However, in reality, there will be a highest peak and a secondary peak that can be compared (comparison: the difference between the peaks of the two sets of data remains within a threshold, and the threshold in this method is 0.1) in the spectrogram at this position, and the situation of two highest peaks will not occur. Therefore, before the respiration frequency detection boundary estimation system, the respiration frequency detection boundary determination condition obtained based on a large amount of experimental data is given first: when the intensity of the respiration signal and the intensity of the noise signal can be compared, the sensing device is on the respiration frequency detection boundary.

[0107] For the respiration signal of personnel and the noise signal to be comparable, it is required that the frequency difference between the highest peak and the secondary peak corresponding in the spectrogram is greater than 0.1 Hz, and the ratio of the difference obtained by subtracting the amplitude of the secondary peak from the amplitude of the highest peak to the amplitude of the secondary peak is less than 0.1. Assuming that the frequencies of the two signals are f1 and f2, and the corresponding amplitudes are Amp1 and Amp2, if the two signals satisfy:

[0108] |f1 - f2| > 0.1 Hz, and

[0109] then it is considered that these two signals can be compared. When the highest peak and the secondary peak in the spectrogram are just comparable, it is considered that this position is the detection boundary.

[0110] The above are only the preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. Any equivalent modifications or changes made by those of ordinary skill in the art according to the disclosure of the present invention shall fall within the protection scope recorded in the claims.

Claims

1. A method for detecting the boundary of human breathing perception based on channel state information, characterized in that: This method includes the following specific steps: Step 1: Collect wireless signal data in the indoor environment under the breathing state of the person, and extract the original channel state information CSI data therefrom; Step 2: Perform data preprocessing on the CSI original data, including outlier removal and noise filtering; Step 3: For the preprocessed data, use subcarrier selection and fast Fourier transform FFT to obtain the breathing frequency of the person; Step 4: Model the processed data, establish an association model between the reflection path length of the wireless signal and the breathing frequency, estimate the reflection path amplitude attenuation and transmission delay, separate the breathing component and the noise component in the CSI amplitude, and derive the breathing frequency detection boundary conditions; In Step 4, it specifically includes the following steps: Step 4-1: Establish the correlation model between the reflection path length d TPR and the breathing frequency f BR ; the reflection path consists of two segments, from the transmitter to the person's abdomen and from the person's abdomen to the receiver Step 4-2: Estimate the amplitude attenuation α TPR and the transmission delay τ TPR ; Obtain the reflection path length from Step 4-1, expand it using the Taylor formula to obtain the amplitude attenuation α TPR and the transmission delay τ TPR ; Step 4-3: Separate the breathing component and the noise component in the CSI amplitude; represent the multipath signal propagation path as the superposition of the direct path and the reflection path; perform Fourier transform on the channel impulse response CIR over a period of time to obtain the channel frequency response CFR during the corresponding period; perform band-pass filtering to retain the signals within the breathing frequency band; Subsequently, the DC component is removed to obtain the respiration frequency and the noise frequency; then, a fast Fourier transform is performed to obtain four frequency points, namely the respiration signal frequency f BR , the noise signal frequency f, the mixed frequency of the respiration and noise signals f + f BR and f - f BR ; the amplitude corresponding to each frequency point is set to AmpZ f , Step 4-4: Solve the boundary d for detecting the respiratory rate of the person RFDB The value specifically includes the following three stages: Stage 1, when the transmitter is relatively close to the abdominal position of the person, at this time the breathing signal is stronger than the noise signal, it is considered that the transmitter is within the breathing frequency detection boundary at this time; Stage 2, when the transmitter is relatively far from the abdominal position of the person, at this time the breathing signal will have a large attenuation, and the breathing signal intensity will be equal to the noise signal intensity, it is considered that the transmitter is on the breathing frequency detection boundary at this time; Stage 3, when the transmitter is very far from the abdominal position of the person, the breathing signal is severely attenuated, and the noise signal is stronger than the breathing signal, it is considered that the transmitter is outside the breathing frequency detection boundary at this time; Amplitude of the respiration signal Is the maximum value among the amplitudes corresponding to 4 frequency points; when the amplitude AmpZ max Of the noise signal is Equal to the amplitude of the respiration signal, it is considered that the transmitter is on the respiration frequency detection boundary RFDB. By modeling and deriving the respiration frequency detection boundary, the condition for judging whether the sensing device is on the respiration frequency detection boundary is obtained, and the theoretically detected boundary d RFDB Of the human respiration frequency is solved; Step 5: Conduct tests based on the model to achieve the detection of the breathing frequency boundary of the person.

2. The method for detecting the boundary of human respiration perception based on channel state information according to claim 1, wherein: In Step 1, a communication device including a transmitter and a receiver is used to collect wireless signal data. The receiver and the transmitter are placed on both sides of the person, and both use the AP mode for wireless communication. The working frequency of WiFi is 2.4 GHz, and the receiver extracts the original channel state information CSI data from the collected WiFi signals.

3. The method for detecting the boundary of human respiration perception based on channel state information according to claim 1, wherein: In Step 2, the data preprocessing includes outlier removal and noise filtering, specifically: Step 2-1: Outlier removal; use the Hampel identifier to remove outliers from the original data; Step 2-2: Noise filtering; use a band-pass filter to filter out noise signals and limit the frequency band of the received signal in the normal breathing frequency band.

4. A method for detecting the boundary of human respiration perception based on channel state information according to claim 1, characterized in that: In Step 3, perform subcarrier selection and fast Fourier transform FFT to obtain the breathing frequency of the person, specifically: Step 3-1: Subcarrier selection method; calculate the variance of the CSI amplitude through a time window, and then select the subcarrier with the largest variance; Step 3-2: Breathing frequency estimation; perform fast Fourier transform FFT on the subcarrier signals selected above to extract the breathing frequency of the experimenter.

5. A method for detecting the boundary of human respiration perception based on channel state information according to claim 1, characterized in that: In step 4, a transmitter Tx and a receiver Rx are defined. The transmitter is placed on one side of the person's body, and the receiver is placed on the other side. Point A is the highest point of the abdomen corresponding to the maximum inhalation of the person, and point B is the lowest point of the abdomen corresponding to the maximum exhalation of the person. Point P is any point between A and B, representing the abdominal position at any moment during the entire breathing process. The transmitted signal is reflected from point P to the receiver. Point C is the projection of point B on the horizontal plane where the transmitter is located. Point D is the projection of point C on the horizontal plane where the receiver is located. The distances between AB, BC, and CD are denoted as d AB , d BC , d CD ; Point E is the intersection of the horizontal line of the receiver Rx and its perpendicular line passing through point D. Point F is the foot of the perpendicular from the straight line passing through the transmitter to the cross-section where the person's abdomen is located. d ER and d TF are the distances from the receiver and the transmitter to the cross-section where the experimenter's abdomen is located, respectively. The distance between FC is denoted as d FC .

6. A method for detecting the boundary of human respiration perception based on channel state information according to claim 5, characterized in that: When d ER is 0, the receiver is directly opposite the experimenter's abdomen; when d ER is not 0, the receiver is not directly opposite the abdomen.

7. A method for detecting the boundary of human respiration perception based on channel state information according to claim 5, characterized in that: When d CD is 0, the receiver and the experimenter are at the same height.

8. A method for detecting the boundary of human respiration perception based on channel state information according to claim 5, characterized in that: In Step 4, it specifically includes the following steps: Step 4-1: Establish the association model between the reflection path length d TPR and the breathing frequency f BR ; the reflection path consists of two segments: from the transmitter to the person's abdomen and from the person's abdomen to the receiver, that is, d TPR The expression consists of d TP and d PR and is composed of two parts. By performing Taylor series expansion on them, the values of d TP and d PR are obtained and denoted as: d TPR = d TP + d PR = N1sin2πft - N2cos4πft + N3 BR t - N2cos4πf BR t + N3 Step 4-2: Estimate the amplitude attenuation α TPR and the transmission delay τ TPR ; Obtain the reflection path length from Step 4-1, expand it using the Taylor formula, and obtain the amplitude attenuation α TPR and the transmission delay τ TPR , as follows: Among them A TPR is a constant, G t , G r are the antenna gains of the transmitter and the receiver respectively, t is time, and λ and f are the propagation wavelength and frequency of the radio signal in the dormitory scenario; Step 4-3: Separate the breathing component and the noise component in the CSI amplitude; represent the multipath signal propagation path as the superposition of the direct path and the reflection path, and the channel impulse response CIR of the indoor wireless multipath channel is expressed as: where α m (t), τ m (t) and θ m (t) are the amplitude attenuation, propagation delay and phase shift of the signal on the m-th path respectively, where m is the TR and TPR, i.e., the direct path and the reflected path; z(t) is the noise; By performing a Fourier transform on h(t) over a period of time, the channel frequency response CFR for the corresponding time period is obtained; H(f) over a period of time is measured, that is, the channel state information CSI. When performing a Fourier transform on consecutive CSI data packets over a time period, it is considered that α m (t), τ m (t) and θ m (f) are approximately invariant and are constants; the CSI measured by one data packet is expressed as: To facilitate the calculation of the CSI amplitude of the received signal, H kΔT (f) is expressed in complex form and simplified. The squared value of the CFR amplitude of the received signal obtained is denoted as: |H kΔT (f)| 2 = α TR 2 + α TPR 2 +|Z kΔT (f)| 2 + 2α TR α TPR cos(2πfτ TPR + θ TPR - 2πfτ TR ) + 2α TR Z R (f)cos2πfτ TR + 2α TPR Z R (f)cos(2πfτ TPR + θ TPR ) - 2α TR Z I (f)sin2πfτ TR - 2α TPR Z I (f)sin(2πfτ TPR + θ TPR ) where Z R (f) and ZI ( f) are the real and imaginary parts of the noise Z(f), j is the imaginary unit symbol, and |·| represents the modulus of a complex number; perform a Taylor series expansion on the above formula and perform band-pass filtering to retain the signal within the breathing frequency band; then remove the DC component to obtain the breathing frequency and the noise frequency; then perform a fast Fourier transform to obtain 4 frequency points, which are the breathing signal frequency f BR , the noise signal frequency f, the mixed frequency of the breathing and noise signals f + f BR and f - f BR ; set the amplitude corresponding to each frequency point to be AmpZf, Step 4-4: Solve the boundary d for detecting the breathing rate of the person RFDB value; During the process of the transmitter gradually moving away from near the person's abdomen, the distance d between the transmitter and the person's abdomen TF gradually increases, resulting in the gradual increase of the reflection path length d TPR gradually increases, and the signal intensity corresponding to the breathing signal will also gradually attenuate, specifically including the following three stages: Stage 1: When the transmitter is close to the abdominal position of the person, the respiratory signal is stronger than the noise signal at this time, and it is considered that the transmitter is within the respiratory frequency detection boundary at this time; Stage 2: When the transmitter is far from the abdominal position of the person, the respiratory signal will be greatly attenuated at this time, and the intensity of the respiratory signal will be equal to the intensity of the noise signal. It is considered that the transmitter is on the respiratory frequency detection boundary at this time; Stage 3: When the transmitter is very far from the abdominal position of the person, the respiratory signal is severely attenuated, and the noise signal is stronger than the respiratory signal. It is considered that the transmitter is outside the respiratory frequency detection boundary at this time; Based on the above analysis, it is expressed as: where AmpZ max is the maximum value in; When inside the RFDB, although the respiration signal will attenuate as d TF increases, it is still stronger than the noise signal. Therefore, the amplitude of the respiration signal is the maximum among the amplitudes corresponding to the 4 frequency points; while there are 3 frequency points corresponding to the noise. Only when the amplitude AmpZ max of the noise signal corresponding to the maximum frequency point is almost equal to the amplitude of the respiration signal, it is considered that the transmitter is on the RFDB. From this, it can be known that the RFDB condition is When this condition is met, the d TF value obtained is the d RFDB value, that is By modeling and deriving the breathing frequency detection boundary, the conditions for judging whether the sensing device is on the breathing frequency detection boundary are obtained, and the theoretically detected boundary d of the human breathing frequency is solved. RFDB value.

9. The method for detecting the boundary of human respiration perception based on channel state information according to claim 1, wherein: In step 5, testing is carried out on the basis of the model proposed in step 4 to achieve the respiratory frequency detection boundary. Specifically: Observe the spectrogram of the data collected by the receiver when the transmitter is in different positions; when the position of the transmitter is just on the detection boundary, there will be a highest peak and a comparable sub-peak in the spectrogram at this position; the determination condition for the respiratory frequency detection boundary given is: when the intensity of the respiratory signal is compared with the intensity of the noise signal, the sensing device is on the respiratory frequency detection boundary; The comparison between the intensity of the person's respiratory signal and the intensity of the noise signal needs to satisfy that the frequency difference corresponding to the highest peak and the sub-peak in the spectrogram is greater than 0.1 Hz, and the ratio of the difference obtained by subtracting the amplitude of the sub-peak from the amplitude of the highest peak to the amplitude of the sub-peak is less than 0.1; assuming that the frequencies of the two signals are f1 and f2, and the corresponding amplitudes are Amp1 and Amp2, if the two signals satisfy: |f1 - f2| > 0.1 Hz, and Then it is considered that the two signals are compared with each other; when the highest peak and the sub-peak in the spectrogram are just compared, it is considered that this position is the detection boundary.

Images