A robust respiration detection method based on wi-fi beamforming

Abstract

The application discloses a kind of robust breath detection methods based on Wi-Fi beam forming, using beam forming algorithm combination multiple antenna on CSI, construct the reference CSI that does not contain breath information, eliminate the interference of time-varying noise by the quotient of two, then use coordinate transformation to highlight breath feature, then use short-time Fourier transform to obtain frequency spectrum, calculate the spatial position of target, extract target breath pattern and breath rate.The method not only can enhance target signal, effectively suppress noise interference, extend breath detection range to the whole room, while mapping breath signal to position-independent feature space to eliminate the perception "blind spot", and further improve robustness.

Description

Technical Field

[0001] This invention belongs to the field of detection technology, specifically relating to a robust respiration detection method. Background Technology

[0002] As one of the most important physiological processes in the human body, respiration can largely reflect a person's health status, making respiration detection particularly important in clinical diagnosis and disease prevention. Existing respiration detection methods can be broadly categorized into contact and non-contact methods. Contact-based methods are relatively mature, but require users to wear sensing devices, which can significantly hinder daily life. Users (such as the elderly) often resist this method, and it cannot handle special scenarios such as burn patients. To address this issue, various non-contact respiration detection methods based on wireless signals (such as radar) have been proposed. However, the high cost of radar equipment severely hinders the promotion and widespread adoption of these methods in home settings.

[0003] In contrast, Wi-Fi devices are inexpensive, widely adopted globally, and generally support MIMO technology, enabling truly contactless detection and seamless access. A 2022 article by PervasiveHealth, "Robust Respiration Sensing Based on Wi-Fi Beamforming," proposed a robust respiration detection method based on Wi-Fi beamforming, which can meet room-level respiration detection needs. However, the main drawback of this method is that it only uses multiple antennas at the receiving end, greatly limiting the detection performance due to the number of receiving antennas. Furthermore, the method still exhibits some location dependence, and user respiration detection still faces robustness issues. Summary of the Invention

[0004] To overcome the shortcomings of existing technologies, this invention provides a robust breathing detection method based on Wi-Fi beamforming. It uses a beamforming algorithm to combine the CSIs (Cyclic Signal Indicators) on multiple antennas, constructs a reference CSI that does not contain breathing information, and quotients the two to eliminate time-varying noise interference. Coordinate transformation is then used to highlight breathing features, followed by short-time Fourier transform to obtain the spectrum, calculate the target's spatial location, and extract the target's breathing pattern and rate. This method not only enhances the target signal and effectively suppresses noise interference, extending the breathing detection range to the entire room, but also maps the breathing signal to a location-independent feature space to eliminate perception "blind spots," thereby improving robustness.

[0005] The technical solution adopted by this invention to solve its technical problem includes the following steps:

[0006] Step 1: Deploy two terminal devices as the transmitter and receiver respectively, collectively referred to as the transceiver, in an indoor environment. Each device is equipped with an antenna array consisting of omnidirectional antennas. The transceiver is 0.8m above the ground.

[0007] Step 2: The computer controls the transmitter to transmit probe data packets, and the receiver receives the probe data packets and calculates the channel state information (CSI) of the channel between all antenna pairs at the transmitter and receiver based on them.

[0008] Step 3: Using the transmitter and receiver as centers, scan different directions in the physical space, calculate the weight vectors, and sum the weighted CSIs of different links to obtain the composite CSI;

[0009] Step 4: Construct a reference CSI that does not contain respiratory information, and divide the synthetic CSI obtained in Step 3 by this reference CSI to eliminate time-varying noise and obtain a denoised CSI;

[0010] Step 5: Perform coordinate transformation on the noise-reduced CSI to enhance the breathing characteristics of the perceived target;

[0011] Step 6: Use short-time Fourier transform to convert the signal obtained in step 5 from the time domain to the frequency domain, extract the direction of the target relative to the transceiver based on the peak information of the spectrum, and then calculate the target's azimuth.

[0012] Step 7: Extract the signal in the peak direction from Step 6, use frequency domain bandpass filtering to extract the target's breathing pattern, and use an autocorrelation-based periodic detection algorithm to calculate the breathing rate, i.e., the number of breaths per minute.

[0013] Furthermore, the antenna array arrangement of the transceiver in step 1 is as follows: the array elements of the antenna array are arranged in a straight line with equal distances d, i.e., a uniform linear array, and the distance between adjacent antennas does not exceed half a wavelength, i.e., less than 6.15cm in the 2.4GHz band and less than 2.83cm in the 5G band.

[0014] Furthermore, in step 2, the rate at which the transmitting end sends probe data packets is set to 100 per second.

[0015] Furthermore, step 3 specifically includes:

[0016] Beamforming uses the antenna array normal direction as 0°, with a scanning angle range of -90° to 90°. The weight vector for CSI weighting is as follows:

[0017] The weight vector for Delayed Summation Beamforming (DSB):

[0018]

[0019] Where d is the antenna spacing, θ is the scanning angle, λ is the signal wavelength, and α is the alignment angle;

[0020] Weight vector of the minimum variance distortion-free response MVDR beamforming algorithm:

[0021]

[0022] in A(θ) is the steering vector, representing the expected received signal.

[0023] Furthermore, in step 4, when constructing the reference CSI, in order to eliminate the respiratory component when the components of the signal are unknown, the respiratory energy ratio, i.e., the ratio of the sum of the energies of each component within a given frequency range to the total energy of the signal, is used as an indicator. First, a genetic algorithm is used to search for an approximate global optimum to solve the problem of potentially getting trapped in a local optimum. Then, the output of the genetic algorithm is used as the initial value, and a stochastic optimization algorithm is used to search for the exact solution of the global optimum to ensure the efficiency of the optimization algorithm.

[0024] Furthermore, step 6 determines the direction of the target relative to the transceiver by scanning, and calculates the position of the target using the triangular relationship between the target, the transmitter, and the receiver.

[0025] The beneficial effects of this invention are as follows:

[0026] This invention utilizes beamforming algorithms to integrate the CSI of multiple antennas, effectively enhancing the target signal, successfully suppressing noise interference, and extending the breathing detection range to the entire room. Simultaneously, beamforming technology is used to construct a reference CSI to eliminate time-varying noise, and coordinate transformation and feature mapping are employed to enhance and extract location-independent breathing features, thereby eliminating perception "blind spots" and significantly improving the system's robustness. Furthermore, by scanning to obtain the target's azimuth information, the approximate location information of the target can be obtained by further utilizing the principle of triangulation. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the equipment deployment for the method of the present invention.

[0028] Figure 2 This is a diagram illustrating the causes of the location-dependent problem involved in the method of this invention.

[0029] Figure 3 This is a diagram illustrating the beamforming principle in the method of the present invention.

[0030] Figure 4 The images show a comparison of the effects of beamforming and beam nulling (a special type of beamforming) on ​​the original CSI using the method of this invention. (a) Original CSI amplitude, (b) CSI amplitude after beamforming, (c) CSI amplitude after beam nulling, (d) Original CSI phase, (e) CSI phase after beamforming, and (f) CSI phase after beam nulling.

[0031] Figure 5 This is the frequency-direction spectrum obtained after short-time Fourier transform in the method of this invention.

[0032] Figure 6 This is a schematic diagram illustrating the principle of locating the target position after extracting the target's azimuth angle relative to the transceiver.

[0033] Figure 7 This is the autocorrelation function curve in a single experiment when the method of the present invention calculates respiratory periodicity based on the autocorrelation coefficient.

[0034] Figure 8 This is a diagram of the respiratory signal pattern obtained by the method of the present invention after bandpass filtering. Detailed Implementation

[0035] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0036] A robust breathing detection method based on Wi-Fi beamforming involves deploying MIMO-enabled transceiver equipment, using beamforming algorithms to combine the CSIs (Cyclic Signal Indicators) on multiple antennas, constructing a reference CSI that does not contain breathing information, quotienting the two to eliminate time-varying noise interference, highlighting breathing features using coordinate transformation, obtaining the spectrum using short-time Fourier transform, calculating the target's spatial location, and extracting the target's breathing pattern and rate. Its main components include signal acquisition and beamforming, feature enhancement, time-frequency transformation and azimuth analysis, and breathing pattern and rate extraction. This invention proposes a robust breathing detection method based on Wi-Fi beamforming. It uses a pair of terminal nodes equipped with Intel 5300 network cards and Linux CSI Tool as transceivers. Through beamforming, noise reduction, coordinate transformation, time-frequency transformation, and spectrum analysis, the azimuth angle of the target relative to the transceivers is obtained. Combined with the spatial location of the transceivers, the spatial location of the target can be calculated from the triangular relationship between the transmitter, receiver, and target. Furthermore, the signal of this azimuth is subjected to bandpass filtering and period detection to extract the breathing pattern of the target and calculate the breathing rate.

[0037] The technical solution of this invention mainly includes the following steps:

[0038] Step 1: Deploy a pair of terminal devices equipped with Intel 5300 network cards, Ubuntu 14.04 systems, and Linux 802.11n CSITool as the transmitter and receiver respectively in a bedroom environment. Each device is equipped with an antenna array consisting of three omnidirectional antennas, each 0.8m above the ground, with a 2m distance between the devices. Figure 1 As shown;

[0039] Step 2: Control the transmitter to transmit probe data packets at a rate of 100 packets / second. The receiver receives the data packets and calculates the Channel State Information (CSI) for the 9 (3×3) links. Figure 4 In the diagram, a) and d) represent the amplitude and phase of the CSI of one of the links, respectively;

[0040] Step 3: Centering on both the transmitter and receiver, with the antenna array normal direction at 0°, scan the planar space within the range of -90° to 90° in 1° increments. Calculate the weight vector and sum the weighted CSI values ​​for different links. Figure 3 To illustrate, a synthetic CSI is obtained. Figure 4 b) and e) show the amplitude and phase of the enhanced signal;

[0041] Step 4: Construct a reference CSI that does not contain breathing information using the beam nulling algorithm, and divide the synthesized CSI obtained in Step 3 by this reference CSI to eliminate the time-varying carrier frequency offset (CFO), sampling frequency offset (SFO), and automatic gain noise (AGN), thus obtaining the denoised CSI, whose amplitude and phase are as follows: Figure 4 As shown in c) and f);

[0042] Step 5: Perform coordinate transformation on the denoised CSI to map the breathing signal to a location-independent feature space, extract breathing features, thereby enhancing the breathing characteristics of the perceived target and solving the location-dependent problem. The cause analysis is as follows: Figure 2 As shown;

[0043] Step 6: Set the time window to 30 seconds, and use a short-time Fourier transform to convert the time-domain signal obtained in Step 5 to the frequency domain, obtaining the target respiration frequency-direction pattern, as shown below. Figure 5 As shown; the direction of the target relative to the transceiver is extracted based on the spectral peak information, and then the target's azimuth is calculated by combining the spatial position of the transceiver. The calculation model is as follows. Figure 6 As shown;

[0044] Step 7: Extract the signal from the peak direction in Step 6, and use a frequency domain bandpass filter to extract the target's breathing pattern, such as... Figure 8 As shown; then, the respiratory rate (in BPM) is calculated based on an autocorrelation-based periodic detection algorithm, and the autocorrelation function in the detection is as follows. Figure 7 As shown.

[0045] Furthermore, the antenna array arrangement at the transceiver end in step 1 is as follows: the array elements of the antenna array are arranged in a uniform linear array with a spacing of 0.0185m (the operating frequency of the device is 5.32GHz, corresponding to a half wavelength of 0.0282m).

[0046] Furthermore, in step 3, beamforming uses the antenna array normal direction as 0° and the scanning angle range as -90° to 90°. The weight vector for CSI weighting is as follows:

[0047] Weight vector for Delayed Summation Beamforming (DSB):

[0048]

[0049] The antenna spacing d is 0.0185m, the scanning angle θ is taken in steps of 1°, the signal wavelength λ is 0.0564m, and the alignment angle α is 0.

[0050] Weight vector of the Minimum Variance Distortion-Free Response (MVDR) beamforming algorithm:

[0051]

[0052] in A(θ) is the steering vector, representing the expected received signal.

[0053] Furthermore, in step 4, when constructing the reference CSI, the respiration ratio is calculated using a frequency range of 10–30 bpm. A genetic algorithm, followed by a stochastic optimization algorithm, is used sequentially to search for approximate global optima and local optima, thus achieving efficient global optima search. Specifically, in the approximate global optima search stage, an initial population of 50 is randomly generated, with a crossover probability set to 0.05, a maximum number of iterations set to 40, and a precision set to 10-10. -5 During the local optimum search phase, the approximate global optimum is used as the initial value and its 2-norm is constrained to 1. The maximum number of iterations is set to 1000 times the number of parameters, and the tolerance is set to 10. -7 .

Claims

1. A robust respiration detection method based on Wi-Fi beamforming, characterized in that, Includes the following steps: Step 1: Deploy two terminal devices as the transmitter and receiver respectively, collectively referred to as the transceiver, in an indoor environment. Each device is equipped with an antenna array consisting of omnidirectional antennas. The transceiver is 0.8m above the ground. Step 2: The computer controls the transmitter to transmit probe data packets, and the receiver receives the probe data packets and calculates the channel state information (CSI) of the channel between all antenna pairs at the transmitter and receiver based on them. Step 3: Using the transmitter and receiver as centers, scan different directions in the physical space, calculate the weight vectors, and sum the weighted CSIs of different links to obtain the composite CSI; Step 4: Construct a reference CSI that does not contain respiratory information using the beam nulling algorithm. Divide the synthetic CSI obtained in Step 3 by this reference CSI to eliminate time-varying noise and obtain the denoised CSI. Step 5: Perform coordinate transformation on the noise-reduced CSI to map the breathing signal to a location-independent feature space, extract breathing features, and thus enhance the breathing features of the perceived target; Step 6: Use short-time Fourier transform to convert the signal obtained in step 5 from the time domain to the frequency domain, extract the direction of the target relative to the transceiver based on the peak information of the spectrum, and then calculate the target's azimuth. Step 7: Extract the signal in the peak direction from Step 6, use frequency domain bandpass filtering to extract the target's breathing pattern, and use an autocorrelation-based periodic detection algorithm to calculate the breathing rate, i.e., the number of breaths per minute; In step 4, when constructing the reference CSI, in order to eliminate the breathing component when the components of the signal are unknown, the breathing energy ratio, i.e., the ratio of the sum of the energies of each component within a given frequency range to the total energy of the signal, is used as an indicator. First, a genetic algorithm is used to search for an approximate global optimum to solve the problem of potentially getting trapped in a local optimum. Then, the output of the genetic algorithm is used as the initial value, and a stochastic optimization algorithm is used to search for the exact solution of the global optimum to ensure the efficiency of the optimization algorithm.

2. The robust respiration detection method based on Wi-Fi beamforming according to claim 1, characterized in that, The antenna array arrangement at the transceiver end in step 1 is as follows: the array elements are spaced at equal intervals. Arranged in a straight line, i.e., a uniform linear array, the distance between adjacent antennas does not exceed half a wavelength, i.e. less than 6.15cm in the 2.4GHz band and less than 2.83cm in the 5GHz band.

3. The robust respiration detection method based on Wi-Fi beamforming according to claim 1, characterized in that, In step 2, the rate at which the transmitting end sends probe data packets is set to 100 per second.

4. A robust respiration detection method based on Wi-Fi beamforming according to claim 1, characterized in that, Step 3 specifically involves: Beamforming is based on the normal direction of the antenna array. The scanning angle range is to The weight vector for CSI weighting: Weight vector of Delayed Summation Beamforming (DSB): ； in, Antenna spacing, For the scanning angle, For the signal wavelength, For the alignment angle; Weight vector of the minimum variance distortion-free response MVDR beamforming algorithm: ； in For the estimation of the expected received signal, This is the guide vector.

5. A robust respiration detection method based on Wi-Fi beamforming according to claim 1, characterized in that, Step 6 determines the direction of the target relative to the transceiver by scanning, and calculates the position of the target using the triangular relationship between the target, the transmitter, and the receiver.

Images