Non-contact human vital signs detection method based on adaptive range gate

Abstract

The present invention discloses a non-contact human vital sign detection method based on an adaptive range gate. The method comprises the following steps: Step A: mixing the echo signal obtained after transmission with the transmission signal to obtain an intermediate frequency signal, and then sampling the intermediate frequency signal to obtain a discrete signal to form a data matrix; Step B: determining the range gate where the target human body is located using a variance-based adaptive range gate selection method to obtain a preprocessed signal; Step C: extracting a human feature signal from the preprocessed signal using a variational pattern extraction algorithm; and Step D: estimating human features from the extracted human feature signal using a fast Fourier transform. The present invention deeply integrates the phase variance-based adaptive range gate selection method and the variational pattern extraction algorithm to achieve accurate estimation of human vital sign detection.

Description

Technical Field

[0001] The present invention belongs to the field of radar signal detection and processing, and in particular relates to a millimeter wave radar non-contact human vital sign detection method. Background Art

[0002] The health and well-being of the human body often requires extensive physiological data to assess. Therefore, physiological parameters such as respiration, heart rate, body temperature, and blood pressure play a crucial role in the biomedical field. Heart rate parameters are an important indicator of the normality of a person's cardiopulmonary activity. Cardiopulmonary activity directly influences the activity of various organs and muscles, and many sudden illnesses often lead to abnormalities in cardiopulmonary activity. Therefore, the detection of heart rate parameters is crucial in fields such as medical monitoring.

[0003] With the advancement of biomedical technology, a growing number of methods are available for measuring heart rate parameters. These methods primarily include electrocardiography, pulse measurement using finger pressure, and phonocardiography. However, these methods are primarily contact-based, with major drawbacks: prolonged wearing of contact-based measuring instruments can cause discomfort; and for some patients with specific conditions, such as those with extensive skin ulcers, contact-based measurement methods can cause secondary damage. Furthermore, in many cases, due to various limitations and other considerations, remote measurement of respiratory and heart rate is necessary, leading to the development of non-contact vital sign detection methods.

[0004] Non-contact vital sign detection technology enables long-distance vital sign detection without contacting the target, and can also perform detection through certain obstacles. Common non-contact detection methods in the vital sign detection field include acoustic vibration detection, optical detection, and infrared detection. While these detection methods can also achieve long-distance vital sign detection without contacting the target, they are also susceptible to environmental influences. With the development of biomedical engineering, radar technology has begun to be applied to vital sign detection, marking a significant advancement in the medical field. Compared to traditional non-contact detection technologies, radar-based non-contact vital sign detection technology has greater penetration and anti-interference capabilities, and can operate 24 hours a day.

[0005] Since radar-based life signal detection technology can achieve long-distance and long-term detection, and the detection process is not affected by environmental factors such as weather and light, the application range of this detection technology is very wide. It is currently mainly used in medical monitoring, post-disaster rescue and other fields.

[0006] There are three main radar systems used in life signal detection: continuous wave (CW) radar, pulsed ultra-wideband (UWB) radar, and linear frequency modulated continuous wave (FMCW) radar. CW radar systems are simple and lack ranging capabilities, making them less resilient to interference and clutter, and unable to distinguish multiple targets. UWB and FMCW radars can suppress some clutter interference through range resolution and distinguish multiple targets at different distances. However, UWB radar has high sampling requirements, and corresponding millimeter wave products are relatively expensive. Therefore, FMCW millimeter wave radar has become the preferred choice for heart rate detection.

[0007] However, current FMCW radar-based contactless heart rate detection still has some problems in selecting the distance gate to the target person and estimating the heartbeat signal spectrum:

[0008] First, even though the target person is stationary, there is still some small body movement. When there is such inevitable random body movement, the previous range gate selection method will not be able to select the range gate that contains the most human vital signs information.

[0009] Secondly, traditional mode decomposition algorithms all have modal aliasing problems to varying degrees and cannot obtain relatively pure vital signs signals (heartbeat signals or respiratory signals). Summary of the Invention

[0010] The present invention proposes a non-contact human vital sign detection method based on an adaptive range gate, the purposes of which are: (1) to solve the problem of being unable to select the range gate that contains the most human vital sign information; and (2) to obtain a purer vital sign signal.

[0011] The technical solutions of the present invention are as follows:

[0012] A non-contact human vital sign detection method based on an adaptive range gate, comprising the following steps:

[0013] Step A: For a human target to be measured, an FMCW radar is selected to transmit N linear frequency modulated pulse train signals to the target human body; each pulse echo signal received by the radar includes the human target echo signal and clutter; the echo signal is mixed with the transmitted signal to obtain an intermediate frequency signal, and the discrete signals obtained by sampling the intermediate frequency signal are then formed into a data matrix, with the results obtained by each pulse sampling being placed in a column;

[0014] Step B: Preprocess the data matrix to obtain a preprocessed signal: A distance matrix is ​​obtained based on the data matrix, and each row of the distance matrix is ​​used as a range gate to filter out static clutter on the distance matrix; the range gate where the human target is located is then determined based on the energy and phase variance to accurately extract the phase information reflected by the human body; the phase is then unwrapped to resolve the breakpoint problem in the phase information, and then phase difference processing is performed on the unwrapped signal to obtain a phase difference signal. The phase difference signal is then subjected to impulse noise removal to obtain a preprocessed signal;

[0015] Step C: extracting human body feature signals from the preprocessed signals using a variational pattern extraction algorithm;

[0016] Step D: Use fast Fourier transform to estimate the human body features of the extracted human body feature signals.

[0017] As a further improvement of the non-contact human vital sign detection method based on the adaptive range gate, step A uses a millimeter wave radar life detection system to obtain the echo signal reflected by the target human body. The millimeter wave radar life detection system includes a signal generator, a power amplifier, a low-noise amplifier, a low-pass filter and an analog-to-digital converter module.

[0018] As a further improvement of the non-contact human vital sign detection method based on the adaptive range gate, step A is specifically as follows:

[0019] Step A.1. Calculate the linear frequency modulation pulse train signal generated by the signal generator at the radar front end using the following formula:

[0020]

[0021] Among them, A T represents the amplitude of the transmitted signal, is the phase noise, B is the bandwidth of the signal, T d is the duration of the signal, f min is the initial frequency of the signal;

[0022] The echo signal from the target human body received by the receiving antenna is expressed as:

[0023]

[0024] Among them, A R represents the amplitude of the echo signal, τ is the time delay of the echo signal relative to the transmitted signal;

[0025] Assuming that the initial distance of the target body relative to the radar is R0, and the chest displacement caused by the signal in the living body when the target body is relatively still is x(t), then the real-time distance between the target body and the radar is R(t)=R0+x(t), and the time delay τ is Where c is the speed of light; the echo signal is amplified by the low-noise amplifier and mixed with the local array signal to generate an intermediate frequency signal expressed as:

[0026]

[0027] in, is the phase of the IF signal;

[0028] In step A.2, the intermediate frequency signal enters the analog-to-digital converter for A / D sampling to obtain a data matrix of the echo signal, and the sampling results of each linear frequency modulation pulse train signal are placed in a column, for a total of N linear frequency modulation signals.

[0029] As a further improvement of the non-contact human vital sign detection method based on the adaptive range gate, the step B specifically includes the following steps:

[0030] Step B.1. Perform an M-dimensional distance-dimensional Fourier transform on each column of the data matrix, for a total of N columns, to obtain a distance matrix R[m,n] containing the target human distance information, where m = 1,…M, n = 1,…N, where m and n represent the row and column numbers of the elements in the matrix, respectively. Each row of the distance matrix R[m,n] is a range gate, and the average value of each range gate is regarded as the static clutter of the range gate. After removing the static clutter, R'[m,n] is obtained. The calculation formula for static clutter filtering is:

[0031]

[0032] As a further improvement of the non-contact human vital sign detection method based on the adaptive range gate, the step B further includes the following steps:

[0033] Step B.2: Divide all frames transmitted by the radar into multiple blocks, each block including K frames; within the same block, use the range gate selected for the first frame for all frames in the block;

[0034] The process of determining the range gate where the human target is located includes: in the first stage, the range gate of the first block is selected based on energy; in the second stage, the range gates of subsequent blocks are selected based on phase variance, and the candidate range gates of each subsequent block only consider the range gates near the range gate selected in the previous block, including the a range gates before and after it, a total of 2a+1 range gates. Among the candidate range gates, the range gate with the largest phase variance between the current frame and the b frames before and after it, a total of 2b+1 frames, is considered to be the range gate where the human target is located; a and b are preset integer values.

[0035] As a further improvement of the non-contact human vital sign detection method based on adaptive range gate, the specific steps of determining the range gate where the human target is located in step B.2 are:

[0036] Let U i represents the i-th block, R j represents the jth range gate, F t represents the tth frame;

[0037] First, the range gates of all frames in the first block are selected based on energy: assuming that among all the range gates in the first frame, the range gate with the maximum energy is R, then all K frames in U1 select R as the range gate where the target person is located;

[0038] Except for the first block, all subsequent blocks select range gates based on variance: for U i , assuming U i-1 The selected range gate is R j , then U i The candidate range gate is [R j-a ,R j+a ]Total 2a+1 range gate, U i The frame for range gate selection is F (i-1)*K+1 , calculated from F (i-1)*K+1-b to F (i-1)*K+1+b The phase variance of a total of 2b+1 frames in the candidate range gate, the range gate with the largest variance is considered as U i The range gate where the target human body is located in all frames;

[0039] The calculation formula for the phase variance at each candidate range gate is:

[0040]

[0041] in, represents the phase of the u-th frame at the v-th range gate, Indicates that from F (i-1)*K+1-b to F (i-1)*K+1+b The average value of the phase at the vth range gate in a total of 2b+1 frames;

[0042] The range gate with the largest phase variance is U i The range gate where the corresponding human target is located.

[0043] As a further improvement of the non-contact human vital sign detection method based on the adaptive range gate, the step B further includes the following steps:

[0044] Step B.3: Extract the phase φ(n) from the range gate selected for each frame and unwrap the phase. The unwrapping formula is as follows:

[0045]

[0046] The unwrapped phase is subjected to phase difference processing, that is, the value of the previous phase is subtracted from the value of the latter phase. The signal after phase difference processing is the phase difference signal; the phase difference signal is subjected to impulse noise removal processing, and the obtained signal is the preprocessed signal s(t).

[0047] As a further improvement of the non-contact human vital sign detection method based on the adaptive range gate, the step C includes the following steps:

[0048] Step C.1: Decompose the preprocessed signal s(t) into the human body feature signal u of the desired mode r (t) and the residual signal f k (t);

[0049] Then, based on the center frequency w of the human body characteristic signal r , the bandwidth I1 of the human body characteristic signal is obtained by minimizing:

[0050]

[0051] in Indicates partial differential with respect to t, δ(t) is Dirac distribution, * is convolution, u r (t) is the human body characteristic signal at time t, i and j are complex units;

[0052] Step C.2: Based on f k (t) and u r The penalty function I2 is defined to minimize the spectral overlap between (t):

[0053]

[0054] Wherein, γ(t) is the impulse response of the preset filter;

[0055] Step C.3: Set human body characteristic signal u r (t), residual signal f k (t) and center frequency w r Satisfaction among the three is the convergence condition, where η is the equilibrium parameter, and then the human characteristic signal u is obtained by iterative sub-optimization r (t), the specific steps are as follows:

[0056] Step C.3.1, the human body feature signal u r (t), residual signal f k (t) and the Lagrange multiplier λ(t) are transformed into the matched filter output represented in the frequency domain through the variational mode Human body characteristic signals residual signal and Lagrange multipliers

[0057] Step C.3.2, Initialization: Let Set loop variable m, set m=1, and start iteration;

[0058] Step C.3.3, at the mth iteration, use the alternating direction method of the multiplier algorithm to calculate and The Lagrange multipliers in the frequency domain are calculated by the dual ascent method

[0059]

[0060]

[0061]

[0062]

[0063] Where τ is the update parameter;

[0064] Step C.3.4, set the discrimination accuracy ζ, and ζ>0,

[0065] like and The convergence between satisfies the following equation:

[0066]

[0067] Then stop the iteration, That is, to obtain the human body characteristic signal represented in the frequency domain

[0068] Otherwise, set m=m+1 and continue with steps C.3.3-C.3.4.

[0069] As a further improvement of the non-contact human vital sign detection method based on the adaptive range gate, the step D includes the following steps:

[0070] Step D.1, perform fast Fourier transform on the human body characteristic signal, and take the frequency f at the maximum spectrum amplitude as the human body characteristic signal frequency;

[0071] Step D.2, calculate the number of human features.

[0072] As a further improvement of the non-contact human vital sign detection method based on the adaptive range gate:

[0073] If the human body characteristic signal is a heartbeat signal, the heartbeat signal frequency obtained in step D.1 is f h, then the number of heart beats per minute is:

[0074] HR = f h ×60;

[0075] If the human body characteristic signal is a breathing signal, the breathing signal frequency obtained in step D.1 is f b , then the number of breaths per minute is:

[0076] BR=f b ×60.

[0077] Compared with the prior art, the present invention has the following beneficial effects:

[0078] The present invention deeply integrates the phase variance-based adaptive range gate selection method and the variational mode extraction algorithm to achieve accurate estimation of human vital signs detection. On the one hand, the present invention can accurately select the range gate where the target human body is located through the variance-based adaptive range gate selection method, and then accurately obtain the target human body's vital signs information, laying the foundation for subsequent vital signs estimation and making high-precision vital signs estimation possible. On the other hand, the present invention uses the variational mode extraction algorithm (VME) to replace the traditional empirical mode decomposition algorithm, and the obtained human characteristic signal is purer, and the amount of calculation is greatly reduced. At the same time, the static clutter filtering technology is also used to eliminate the influence of static clutter, thereby more clearly displaying the range gate where the target human body is located, and achieving accurate detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0079] Figure 1 Schematic diagram of the overall process of the method of the present invention;

[0080] Figure 2 Schematic diagram of the result of range-dimensional Fourier transform (Range-FFT);

[0081] Figure 3 Figure 1 is a comparison diagram before and after static clutter filtering; Figure a is the result diagram before static clutter filtering, and Figure b is the result diagram after static clutter filtering;

[0082] Figure 4 Select a schematic diagram for the range gate;

[0083] Figure 5 Spectrum diagram of phase signal extracted from range gates selected by two different methods;

[0084] Figure 6 These are the estimated and reference values ​​of heart rate in Experiment 6. DETAILED DESCRIPTION

[0085] The technical solution of the present invention is described in detail below with reference to the accompanying drawings:

[0086] like Figure 1 , a non-contact human vital sign detection method based on adaptive range gate, the steps are:

[0087] Step A: For a human target to be measured, an FMCW radar is selected to transmit N linear frequency modulated pulse train signals to the target human body; each pulse echo signal received by the radar includes the human target echo signal and clutter; the echo signal is mixed with the transmitted signal to obtain an intermediate frequency signal, and the discrete signals obtained by sampling the intermediate frequency signal are then formed into a data matrix, with the results obtained from each pulse sampling being placed in a column.

[0088] In this embodiment, we use the Texas Instruments millimeter wave AWR 1642 radar operating at 77-81 GHz. The millimeter wave radar life detection system includes a signal generator, a power amplifier, a low-noise amplifier, a low-pass filter, and an analog-to-digital converter module.

[0089] Table 1 Main parameters of radar system

[0090] parameter Numerical Initial frequency 77GHz Frequency bandwidth 3.99GHz Frequency slope 70MHz Number of samples per frame 200 ADC sampling time 50μs ADC initial time 7μs Frame interval 50ms .

[0091] In order to verify the effectiveness of our proposed invention, we used the DCA 1000 acquisition board to collect data and transferred the data to the computer terminal through the USB interface, and then used MATLAB for simulation and analysis.

[0092] The subjects wore a Polar H10 chest heart rate sensor and sat 1 meter away from the radar, with their chest at the same level as the radar. The subjects kept their breathing and heart rate stable before the test and remained as still as possible during the test, with no other human targets within the sensor's range.

[0093] Step A specifically includes:

[0094] Step A.1. Calculate the linear frequency modulation pulse train signal generated by the signal generator at the radar front end using the following formula:

[0095]

[0096] Among them, A T represents the amplitude of the transmitted signal, is the phase noise, B is the bandwidth of the signal, T d is the duration of the signal, f mm is the initial frequency of the signal;

[0097] The echo signal from the target human body received by the receiving antenna is expressed as:

[0098]

[0099] Among them, A R represents the amplitude of the echo signal, and τ is the time delay of the echo signal relative to the transmitted signal.

[0100] Assuming that the initial distance of the target body relative to the radar is R0, and the chest displacement caused by the signal in the living body when the target body is relatively still is x(t), then the real-time distance between the target body and the radar is R(t)=R0+x(t), and the time delay τ is Where c is the speed of light; the echo signal is amplified by the low-noise amplifier and mixed with the local array signal to generate an intermediate frequency signal expressed as:

[0101]

[0102] in, is the phase of the IF signal.

[0103] In step A.2, the intermediate frequency signal enters the analog-to-digital converter for A / D sampling to obtain a data matrix of the echo signal, and the sampling results of each linear frequency modulation pulse train signal are placed in a column, for a total of N linear frequency modulation signals.

[0104] Step B: Preprocess the data matrix to obtain a preprocessed signal: obtain a distance matrix based on the data matrix, and use each row of the distance matrix as a range gate to filter out static clutter on the distance matrix; then determine the range gate where the human target is located based on the energy and phase variance to accurately extract the phase information reflected by the human body; then unwrap the phase to resolve the breakpoint problem in the phase information, and then perform phase difference processing on the unwrapped signal to obtain a phase difference signal, and then remove impulse noise from the phase difference signal to obtain a preprocessed signal.

[0105] The step B specifically comprises the following steps:

[0106] Step B.1. Perform an M-dimensional Range-FFT on each column of the data matrix, for a total of N columns, to obtain the distance matrix R[m,n] containing the target human distance information, where m = 1, ...M, n = 1, ...N, where m and n represent the row and column numbers of the elements in the matrix, respectively. Each row of the distance matrix R[m,n] is a range gate, and the average value of each range gate is considered the static clutter of that range gate. After removing the static clutter, R'[m,n] is obtained. The calculation formula for static clutter filtering is:

[0107]

[0108] The process of Range-FFT is as follows Figure 2 The result before static clutter filtering is shown as Figure 3 As shown in a, the result after static clutter filtering is as follows Figure 3As shown in Figure b, there are many false peaks before static clutter filtering, but after static filtering, the location of the target human body is more clearly displayed.

[0109] Step B.2: All frames transmitted by the radar are divided into multiple blocks, each block includes K frames; within the same block, the range gate selected for the first frame is used for all frames in the block.

[0110] The process of determining the range gate where the human target is located includes: in the first stage, the range gate of the first block is selected based on energy; in the second stage, the range gates of subsequent blocks are selected based on phase variance, and the candidate range gates of each subsequent block only consider the range gates near the range gate selected in the previous block, including the a range gates before and after it, a total of 2a+1 range gates. Among the candidate range gates, the range gate with the largest phase variance between the current frame and the b frames before and after it, a total of 2b+1 frames, is considered to be the range gate where the human target is located; a and b are preset integer values.

[0111] The specific steps of determining the range gate where the human target is located in step B.2 are:

[0112] Let U i represents the i-th block, R j represents the jth range gate, F t represents the tth frame;

[0113] First, the range gates of all frames in the first block are selected based on energy: assuming that among all the range gates in the first frame, the range gate with the maximum energy is R, then all K frames in U1 select R as the range gate where the target person is located.

[0114] Except for the first block, all subsequent blocks select the range gate based on the variance: for U i , assuming U i-1 The selected range gate is R j , then U i The candidate range gate is [R j-a ,R j+a ]Total 2a+1 range gate, U i The frame for range gate selection is F (i-1)*K+1 , calculated from F (i-1)*K+1-b to F (i-1)*K+1+b The phase variance of a total of 2b+1 frames in the candidate range gate, the range gate with the largest variance is considered as U i The range gate where the target person is located in all frames.

[0115] The calculation formula for the phase variance at each candidate range gate is:

[0116]

[0117] in, represents the phase of the u-th frame at the v-th range gate, Indicates that from F (i-1)*K+1-b to F (i-1)*K+1+b The average phase value of the vth range gate in 2b+1 frames.

[0118] The range gate selection process for all frames in the i-th block is as follows Figure 4 As shown, the range gate with the largest phase variance is U i The range gate where the corresponding human target is located.

[0119] Step B.3: Extract the phase φ(n) from the range gate selected for each frame and unwrap the phase. The unwrapping formula is as follows:

[0120]

[0121] The unwrapped phase is subjected to phase difference processing, that is, the value of the previous phase is subtracted from the value of the latter phase. The signal after phase difference processing is the phase difference signal; the phase difference signal is subjected to impulse noise removal processing, and the obtained signal is the preprocessed signal s(t).

[0122] Step C: Use the variational pattern extraction algorithm (VME) to extract the human body feature signal from the preprocessed signal.

[0123] Specifically, the step C includes the following steps:

[0124] Step C.1: Decompose the preprocessed signal s(t) into the human body feature signal u of the desired mode r (t) and the residual signal f k (t).

[0125] Then, based on the center frequency w of the human body characteristic signal r (For example, the frequency range of the human heartbeat signal is 0.8-2Hz, and the heart rate is generally more than 70 bpm, so the center frequency is set to 1.2Hz). The bandwidth I1 of the human characteristic signal is obtained by minimizing:

[0126]

[0127] in Indicates partial differential with respect to t, δ(t) is Dirac distribution, * is convolution, u r (t) is the human body characteristic signal at time t, and i and k are complex units.

[0128] Step C.2: Based on f k (t) and u r The penalty function I2 is defined to minimize the spectral overlap between (t):

[0129]

[0130] Wherein, γ(t) is the impulse response of the preset filter.

[0131] Step C.3: Set human body characteristic signal u r (t), residual signal f k (t) and center frequency w r Satisfaction among the three is the convergence condition, where η is the equilibrium parameter, and then the human body characteristic signal u is obtained by iterative sub-optimization r (t). The specific steps are as follows:

[0132] Step C.3.1, the human body feature signal u r (t), residual signal f k (t) and the Lagrange multiplier λ(t) are transformed into the matched filter output represented in the frequency domain through the variational mode Human body characteristic signals residual signal and Lagrange multipliers

[0133] Step C.3.2, Initialization: Let Set the loop variable m, set m=1, and start the iteration.

[0134] Step C.3.3, at the mth iteration, use the alternating direction method of the multiplier algorithm to calculate and The Lagrange multipliers in the frequency domain are calculated by the dual ascent method

[0135]

[0136]

[0137]

[0138]

[0139] Where τ is the update parameter.

[0140] Step C.3.4, set the discrimination accuracy ζ, and ζ>0,

[0141] like and The convergence between satisfies the following equation:

[0142]

[0143] Then stop the iteration, That is, to obtain the human body characteristic signal represented in the frequency domain

[0144] Otherwise, set m=m+1 and continue with steps C.3.3 to C.3.4.

[0145] Table 2 VME parameters

[0146] parameter Numerical Center frequency 1.2 Penalty Factor 20000 Noise tolerance 0 Convergence tolerance 1e-7 Data fidelity balance parameter 20000 .

[0147] Step D: Use fast Fourier transform to estimate the human body features of the extracted human body feature signals. Specifically, the following steps are included:

[0148] Step D.1, perform fast Fourier transform on the human body characteristic signal, and take the frequency f at the maximum spectrum amplitude as the human body characteristic signal frequency;

[0149] Step D.2, calculate the number of human body features.

[0150] If the human body characteristic signal is a heartbeat signal, the heartbeat signal frequency obtained in step D.1 is f h , then the number of heart beats per minute is:

[0151] HR = f h ×60.

[0152] If the human body characteristic signal is a breathing signal, the breathing signal frequency obtained in step D.1 is f b , then the number of breaths per minute is:

[0153] BR=f b ×60.

[0154] In this embodiment, the interference of static clutter is removed through static clutter removal processing, so that the range gate where the target human body is located is displayed more clearly. This method accurately selects the range gate containing human vital signs information by adopting a variance-based adaptive range gate selection method. Considering that when the human body is relatively still, the range gate changes very little, so the subsequent range gate selection only considers the range gate near the previously selected range gate, reducing the amount of calculation. In addition, in order to reduce the interference of noise, the phase change of the current frame and its nearby frames is also considered. This method also extracts the heartbeat signal through VME, which greatly reduces the phenomenon of frequency aliasing. Finally, the obtained heartbeat signal is measured by FFT, and the measured frequency is used as the final heart rate (respiration rate detection is similar to this).

[0155] Experimental verification

[0156] In order to verify the technical solution of the present invention, the following experiment is set up to detect the human heart rate:

[0157] In this experiment, six groups of experiments were set up under the condition of normal heartbeat of the subjects to verify the superiority of the variance-based adaptive range gate selection method. Other test equipment parameters and environment were consistent.

[0158] Experiment 1: Use the traditional energy-based range gate selection method to select a fixed range gate and extract the phase signal from the selected range gate.

[0159] Experiment 2: A variance-based adaptive range gate selection method is used to select a range gate every K frames, and the phase signal is extracted from the selected range gate.

[0160] Experiment 3: The phase signal extracted in Experiment 1 was used to extract the heartbeat signal using VME, and the obtained heartbeat signal was measured using FFT.

[0161] Experiment 4: We used the traditional variance-based range gate selection method to select a fixed range gate and extract the phase signal from the selected range gate. The extracted phase signal was used to extract the heartbeat signal using VME, and the obtained heartbeat signal was measured using FFT.

[0162] Experiment 5: Every K frames, a range gate is selected based on energy, and the phase signal is extracted from the selected range gate. The extracted phase signal is used to extract the heartbeat signal using VME, and the obtained heartbeat signal is measured using FFT.

[0163] Experiment 6: The phase signal extracted in Experiment 2 was used to extract the heartbeat signal using VME, and the obtained heartbeat signal was measured using FFT.

[0164] like Figure 5 As shown in Figure 1, the frequency spectra obtained by performing FFT on the phase signals extracted from Experiment 1 and Experiment 2 are shown. The dotted line is the frequency spectra obtained by performing FFT on the phase signal obtained from Experiment 1, and the solid line is the frequency spectra obtained by performing FFT on the phase signal obtained from Experiment 2. The reference heart rate given by the Polar H10 chest strap sensor is 79 bpm, and the corresponding heart rate signal frequency is 1.316 Hz. Figure 5 As shown in the figure, the peak frequency of the solid line is 1.317 Hz, while the peak in the dotted line is offset. The actual heartbeat peak is masked by the noise signal. Therefore, the range gate selected by the variance-based adaptive range gate contains more heartbeat information and has less interference.

[0165] To evaluate the accuracy of the heart rate estimation of our solution, we use the heart rate measured by a Polar H10 chest strap sensor as the reference heart rate and use the mean absolute error (MAE) as the test metric, defined as:

[0166]

[0167] in:

[0168] W represents the total number of time windows within the observation time;

[0169] BPM true (l) represents the reference value in the lth time window;

[0170] BPM est (l) represents the measurement value within l time windows.

[0171] To mitigate the potential for randomness in experimental results, we selected 10 individuals of varying ages and genders and conducted multiple experiments using the three experimental methods, calculating the average values. The mean absolute errors of the heart rates of these 10 volunteers measured at rest using Experiments 3, 4, 5, and 6 were 2.219 bpm, 2.188 bpm, 3.234 bpm, and 1.453 bpm, respectively. This demonstrates that the variance-based adaptive range gate selection method proposed in this paper can improve heart rate estimation accuracy, demonstrating the effectiveness of this range gate selection method.

[0172] To further verify the stability of the proposed algorithm, we performed a comparative analysis of the stability by calculating the standard deviation of MAE of 10 different data sets, as shown in Table 3.

[0173] Table 3 MAE of heart rate measured in different experiments at rest

[0174]

[0175] At a distance of 1 meter from the radar, the standard deviation of the MAE of the heart rate detection method in Experiment 6 at rest is 0.445, while the standard deviations of the MAE of the heart rate detection methods in Experiments 3, 4, and 5 are 0.848, 0.887, and 1.669, respectively. The results show that the heart rate detection under the present invention has high stability.

[0176] Figure 6 This figure shows a comparison of the heart rate values ​​obtained in Experiment 6 at rest with the heart rate reference values ​​provided by a Polar H10 chest strap sensor. It can be seen that the estimated heart rate of the present invention is almost identical to the reference heart rate value, with the same fluctuation trend, and can well reflect the current heart rate fluctuation.

Claims

1. A non-contact human vital sign detection method based on adaptive range gate, characterized in that The steps are: Step A: For a human target to be measured, an FMCW radar is selected to transmit N linear frequency modulated pulse train signals to the target human body; each pulse echo signal received by the radar includes the human target echo signal and clutter; the echo signal is mixed with the transmitted signal to obtain an intermediate frequency signal, and the discrete signals obtained by sampling the intermediate frequency signal are then formed into a data matrix, with the results obtained by each pulse sampling being placed in a column; Step B: Preprocess the data matrix to obtain a preprocessed signal: A distance matrix is ​​obtained based on the data matrix, and each row of the distance matrix is ​​used as a range gate to filter out static clutter on the distance matrix; the range gate where the human target is located is then determined based on the energy and phase variance to accurately extract the phase information reflected by the human body; the phase is then unwrapped to resolve the breakpoint problem in the phase information, and then phase difference processing is performed on the unwrapped signal to obtain a phase difference signal. The phase difference signal is then subjected to impulse noise removal to obtain a preprocessed signal; Step C: extracting human body feature signals from the preprocessed signals using a variational pattern extraction algorithm; Step D: performing human body feature estimation on the extracted human body feature signals using a fast Fourier transform.

2. The non-contact human vital sign detection method based on adaptive range gate according to claim 1, characterized in that: Step A uses a millimeter wave radar life detection system to obtain an echo signal reflected by a target human body. The millimeter wave radar life detection system includes a signal generator, a power amplifier, a low-noise amplifier, a low-pass filter, and an analog-to-digital converter module.

3. The non-contact human vital sign detection method based on adaptive range gate according to claim 1, characterized in that Step A is specifically as follows: Step A.

1. Calculate the linear frequency modulation pulse train signal generated by the signal generator at the radar front end using the following formula: Among them, A T represents the amplitude of the transmitted signal, is the phase noise, B is the bandwidth of the signal, T d is the duration of the signal, f min is the initial frequency of the signal; The echo signal from the target human body received by the receiving antenna is expressed as: Among them, A R represents the amplitude of the echo signal, and τ is the time delay of the echo signal relative to the transmitted signal. Assuming that the initial distance of the target body relative to the radar is R0, and the chest displacement caused by the signal in the living body when the target body is relatively still is x(t), then the real-time distance between the target body and the radar is R(t) = R0 + x(t), and the time delay τ is Where c is the speed of light; the echo signal is amplified by the low-noise amplifier and mixed with the local array signal to generate an intermediate frequency signal expressed as: in, is the phase of the IF signal; In step A.2, the intermediate frequency signal enters the analog-to-digital converter for A / D sampling to obtain a data matrix of the echo signal, and the sampling results of each linear frequency modulation pulse train signal are placed in a column, for a total of N linear frequency modulation signals.

4. The non-contact human vital sign detection method based on adaptive range gate according to claim 1, characterized in that The step B specifically comprises the following steps: Step B.

1. Perform an M-dimensional distance-dimensional Fourier transform on each column of the data matrix, for a total of N columns, to obtain a distance matrix R[m,n] containing the target human distance information, where m = 1,…M, n = 1,…N, where m and n represent the row and column numbers of the elements in the matrix, respectively. Each row of the distance matrix R[m,n] is a range gate, and the average value of each range gate is regarded as the static clutter of the range gate. After removing the static clutter, R'[m,n] is obtained. The calculation formula for static clutter filtering is:

5. The non-contact human vital sign detection method based on adaptive range gate according to claim 4, characterized in that The step B further comprises the following steps: Step B.2: Divide all frames transmitted by the radar into multiple blocks, each block including K frames; within the same block, use the range gate selected for the first frame for all frames in the block; The process of determining the range gate where the human target is located includes: in the first stage, the range gate of the first block is selected based on energy; in the second stage, the range gates of subsequent blocks are selected based on phase variance, and the candidate range gates of each subsequent block only consider the range gates near the range gate selected in the previous block, including the a range gates before and after it, a total of 2a+1 range gates. Among the candidate range gates, the range gate with the largest phase variance between the current frame and the b frames before and after it, a total of 2b+1 frames, is considered to be the range gate where the human target is located; a and b are preset integer values.

6. The non-contact human vital sign detection method based on adaptive range gate according to claim 5, characterized in that The specific steps of determining the range gate where the human target is located in step B.2 are: Let U i represents the i-th block, R j represents the jth range gate, F t represents the tth frame; First, the range gates of all frames in the first block are selected based on energy: assuming that among all the range gates in the first frame, the range gate with the maximum energy is R, then all K frames in U1 select R as the range gate where the target person is located; Except for the first block, all subsequent blocks select the range gate based on the variance: for U i , assuming U i-1 The selected range gate is R j , then U i The candidate range gate is [R j-a ,R j+a ]Total 2a+1 range gate, U i The frame for range gate selection is F (i-1)*K+1 , calculated from F (i-1)*K+1-b to F (i-1)*K+1+b The phase variance of a total of 2b+1 frames in the candidate range gate, the range gate with the largest variance is considered as U i The range gate where the target human body is located in all frames; The calculation formula for the phase variance at each candidate range gate is: in, represents the phase of the u-th frame at the v-th range gate, Indicates that from F (i-1)*K+1-b to F (i-1)*K+1+b The average value of the phase at the vth range gate in a total of 2b+1 frames; The range gate with the largest phase variance is U i The range gate where the corresponding human target is located.

7. The non-contact human vital sign detection method based on adaptive range gate according to claim 5, characterized in that The step B further comprises the following steps: Step B.3: Extract the phase φn from the range gate selected for each frame and unwrap the phase. The unwrapping formula is as follows: The unwrapped phase is subjected to phase difference processing, that is, the value of the previous phase is subtracted from the value of the latter phase. The signal after phase difference processing is the phase difference signal; the phase difference signal is subjected to impulse noise removal processing, and the obtained signal is the preprocessed signal s(t).

8. The non-contact human vital sign detection method based on adaptive range gate according to claim 5, characterized in that Described step C comprises the following steps: Step C.1: Decompose the preprocessed signal s(t) into the human body feature signal u of the desired mode r (t) and the residual signal f k (t); Then, based on the center frequency w of the human body characteristic signal r , the bandwidth I1 of the human body characteristic signal is obtained by minimizing: in Indicates partial differential with respect to t, δ(t) is Dirac distribution, * is convolution, u r (t) is the human body characteristic signal at time t, i and j are complex units; Step C.2: Based on f k (t) and u r The penalty function I2 is defined to minimize the spectral overlap between (t): Wherein, γ(t) is the impulse response of the preset filter; Step C.3: Set human body characteristic signal u r (t), residual signal f k (t) and center frequency w r Satisfaction among the three is the convergence condition, where η is the equilibrium parameter, and then the human characteristic signal u is obtained by iterative sub-optimization r (t), the specific steps are as follows: Step C.3.1, the human body feature signal u r (t), residual signal f k (t) and the Lagrange multiplier λ(t) are transformed into the matched filter output represented in the frequency domain through the variational mode Human body characteristic signals residual signal and Lagrange multipliers Step C.3.2, Initialization: Let Set loop variable m, set m=1, and start iteration; Step C.3.3, at the mth iteration, use the alternating direction method of the multiplier algorithm to calculate and The Lagrange multipliers in the frequency domain are calculated by the dual ascent method Where τ is the update parameter; Step C.3.4, set the discrimination accuracy ζ, and ζ>0, like and The convergence between satisfies the following equation: Then stop the iteration, That is, to obtain the human body characteristic signal represented in the frequency domain Otherwise, set m=m+1 and continue with steps C.3.3-C.3.

4.

9. The non-contact human vital sign detection method based on an adaptive range gate according to any one of claims 1 to 7, characterized in that Described step D comprises the following steps: Step D.1, perform fast Fourier transform on the human body characteristic signal, and take the frequency f at the maximum spectrum amplitude as the human body characteristic signal frequency; Step D.2, calculate the number of human body features.

10. The non-contact human vital sign detection method based on adaptive range gate according to claim 9, characterized in that: If the human body characteristic signal is a heartbeat signal, the heartbeat signal frequency obtained in step D.1 is f h , then the number of heart beats per minute is: HR=f h ×60; If the human body characteristic signal is a breathing signal, the breathing signal frequency obtained in step D.1 is f b , then the number of breaths per minute is: BR=f b ×60。

Images