Vital Sign Detection Method Based on Spatiotemporal Equivalent Virtual Array Technology

Abstract

A vital sign detection method based on spatio-temporal equivalent virtual array technology. According to the phase information demodulated from the received single-channel continuous wave radar signal, an equivalent multi-channel virtual array signal is obtained through signal recombination. Then, the subspace method is used to perform frequency domain estimation on the multi-channel virtual array signal to obtain the incoming wave angle information. According to the target frequency converted from the incoming wave angle information, the vital sign information of the target is accurately obtained through a coarse-to-fine frequency estimation method. The present invention utilizes the similarity between the single-channel phase signal of a CW radar with single input and single output (SISO) and the multi-channel array signal of a radar with single input and multiple outputs (SIMO) to synthesize an equivalent virtual array, complete the transformation from time to space, and thus achieve accurate and stable estimation of the breathing frequency and heart rate of the human body with super resolution.

Description

Technical Field

[0001] The present invention relates to a technology in the field of medical monitoring, specifically a method for detecting vital signs based on spatio-temporal equivalent virtual array technology. Background Art

[0002] Both the breathing and heartbeat of the human body will cause very small displacements on the surface of the chest cavity. These displacements will generate the Doppler effect. By using a continuous wave radar to transmit electromagnetic waves and analyzing the echo signal, the displacement information on the surface of the human chest cavity can be effectively restored. The detection of vital signs based on radar, such as obtaining the breathing rate and heart rate, can be expressed as a spectrum estimation problem. Currently, the most popular methods are frequency domain extraction algorithms led by the discrete Fourier transform (DFT) and time domain feature extraction methods. However, due to spectral leakage caused by the length of the sampled data and limited resolution limitations, the accuracy of the DFT algorithm is greatly reduced. In addition, since the displacement amplitude caused by breathing is much stronger than that caused by the heartbeat, the heartbeat signal is easily submerged by the third or fourth harmonic of breathing in the spectrum. Summary of the Invention

[0003] Aiming at the problems of insufficient accuracy and resolution of frequency estimation and high computational complexity for heart rate and breathing rate in the prior art, as well as the deficiencies that existing MIMO radars cannot obtain high-precision positioning information for indoor multi-target vital sign detection and are easily affected by surrounding clutter interference resulting in inaccurate positioning, the present invention proposes a method for detecting vital signs based on spatio-temporal equivalent virtual array technology. By using the similarity between the single-channel phase signal of a single-input single-output (SISO) CW radar and the multi-channel array signal of a single-input multi-output (SIMO) radar, an equivalent virtual array is synthesized to complete the transformation from time to space, thereby achieving accurate and stable estimation of the breathing frequency and heart rate of the human body with super resolution.

[0004] The present invention is achieved through the following technical solutions:

[0005] The present invention relates to a method for detecting vital signs based on spatio-temporal equivalent virtual array technology. According to the phase information demodulated from the received single-channel continuous wave radar signal, an equivalent multi-channel virtual array signal is obtained through signal recombination; then, the subspace method is used to perform frequency domain estimation on the multi-channel virtual array signal to obtain the arrival angle information of the incoming wave; according to the target frequency converted from the arrival angle information, the vital sign information of the target is accurately obtained through a frequency estimation method from coarse to fine.

[0006] The demodulation mentioned above refers to: for the received single-channel continuous wave radar signal, first mix the received signal with the transmitted signal for down-conversion, and after obtaining two orthogonal intermediate frequency signals, perform circular fitting on the I and Q signals to complete amplitude and offset calibration; then use the MDACM algorithm for phase demodulation to obtain the phase information of the target.

[0007] The signal recombination mentioned above refers to: cutting the single-channel phase information in the time dimension, specifically: decomposing the phase information matrix Φ(t) at time t into M sub-signals with a length of K, where: the i-th sub-signal is χ i (t) = Φ(t - (i - 1)Kt s ), t s = 1 / f s is the sampling time interval, f s is the sampling rate, and each sub-signal is an equivalent virtual array, that is, an element of an equivalent SIMO radar.

[0008] The subspace method mentioned above refers to: using the subspace estimation (MUSIC) algorithm and the ESPRIT algorithm in the theory of incident angle estimation to estimate the received signal matrix composed of M elements of the SIMO radar to obtain the incident angle information.

[0009] The conversion mentioned above refers to: for the incident angle information obtained by the subspace method, it is equivalent to the frequency values of the respective frequency components in the phase signal of the human chest cavity, specifically: θ i = sin(2kf i / f s ) -1 , where: θ i is the i-th angle information of the incident angle, f i is the frequency value of the i-th component in the phase information of the human chest cavity, f s is the radar signal sampling rate.

[0010] The frequency estimation method from coarse to fine mentioned above refers to: during signal recombination, according to the MUSIC method, select a smaller length K of the sub-signal, and according to the heart rate value obtained by coarse estimation, optimize the length K of the sub-signal with the ESPRIT method until the optimal selection value of the length K of each sub-signal corresponding to the frequency component f i to be estimated currently is obtained, so as to achieve accurate heart rate and respiratory rate estimation results.

[0011] Technical effects

[0012] The present invention synthesizes an equivalent virtual array by utilizing the similarity between the single-channel phase signal of a SISO CW radar and the multi-channel array signals of a SIMO radar, completes the equivalent signal cutting and recombination from time to space, and uses the different frequency components of the human chest surface movement and the equivalent theory of different arrival direction angles of electromagnetic waves incident on the virtual array to complete the extraction of the respiration rate and heart rate using a subspace-based DOA estimation algorithm. By using a coarse-to-fine heart rate estimation method, real-time and accurate heart rate extraction is achieved. The present invention is verified on a 24GHz single-transmission and single-reception CW radar platform and can obtain the heart rate value of the human body with super-resolution accuracy. In more than 10 sitting-state experiments lasting for ten minutes, the RMSE of the heart rate estimation obtained by this method is 0.81 beats per minute compared with the result measured by the gold standard PPG, and the accuracy is improved by more than 12% compared with the frequency estimation algorithm based on DFT. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic diagram of the system of the present invention;

[0014] Figure 2 It is a flowchart of this method;

[0015] Figure 3 It is a schematic diagram of the time-space equivalence principle of the present invention;

[0016] Figure 4 It is a schematic diagram of the coarse-to-fine frequency estimation method of the present invention;

[0017] Figure 5 It is a scene diagram of the embodiment;

[0018] Figure 6 It is a comparison diagram of the spectrum results of this embodiment;

[0019] Figure 7 It is a comparison diagram of the 10-minute heart rate estimation results of this embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] As Figure 1As shown in the figure, this embodiment relates to a vital sign detection system based on spatio-temporal equivalent virtual array technology, including: a signal source, a power amplifier, a transmitting antenna and a receiving antenna, a low-noise amplifier, a mixer, an operational amplifier, an MCU with ADC and DAC functions, and a backend signal processing unit. Among them: The signal generated by the signal source passes through the power amplifier and is then transmitted by the radar transmitting antenna unit; the transmitted electromagnetic wave is reflected by the target to be measured and received by the radar antenna receiving unit. The received signal is amplified by the low-noise amplifier and down-converted by the mixer to obtain an intermediate-frequency signal. The analog-to-digital conversion unit in the MCU samples the intermediate-frequency signal, and the voltage output by the DAC is fed back to the amplifier to ensure that the amplifier operates in the optimal working area. The backend signal processing unit sequentially reorganizes the sampled signal in the time dimension to obtain a virtual multi-channel signal, performs subspace method to complete the DOA estimation of the arrival angle of the incoming wave, and uses the time-space equivalent principle to complete the conversion between the arrival angle of the incoming wave and the frequency of the target signal to obtain the respiration rate and heart rate information of the target to be measured, and finally realizes the detection of the respiration rate and heart rate of the target.

[0021] As Figure 2 shown, the reorganization in the time dimension mentioned above refers to: cutting the received single-channel radar phase signal in the time dimension. Specifically: the size of the phase information matrix Φ(t) at time t in the time dimension is 1*MK, which is composed of the phase history information of the past MK moments; Φ(t) is decomposed into M sub-signals with a length of K, and the splitting principle of the i-th sub-signal follows χ i (t) = Φ(t - (i - 1)Kt s ), where: t s = 1 / f s is the sampling time interval, f s is the sampling rate, and each sub-signal can be regarded as an equivalent virtual array, that is, an element of an equivalent SIMO radar.

[0022] When the received single-channel radar phase signal whose size is 1*MK, after splitting, it generates M sub-signals χ(t) = [χ1(t), χ2(t),..., χ M (t0], where the size of each sub-signal is 1*K, and the i-th sub-signal matrix can So far, the conversion from the single-channel SISO radar signal to the multi-channel SIMO radar virtual channel signal is completed.

[0023] As Figure 3As shown, the conversion between the incoming wave angle and the frequency of the target signal means that the phase signal d(t) on the surface of the human chest can be caused by the combined displacement due to the breathing motion of the lungs, the contraction motion of the atria and ventricles, and their harmonics, that is where: P represents the number of all frequency components, A i , ω i , respectively represent the amplitude information, angular frequency information, and initial phase of the i-th frequency component. When the phase information demodulated from the CW radar intermediate frequency signal is Φ(t) = d(t) + n(t), where: n(t) is the additive Gaussian white noise caused by the residual phase, environmental interference, and circuit noise.

[0024] After the signal recombination in the time dimension as shown in Figure 2 , the overall signal where: A(ω) = [a(ω1), a(ω2),..., a(ω P )] is the array manifold matrix, is the signal steering vector of ω i , S(t) = [s1(t), s2(t),..., s P (t)] T is the signal matrix, where the i-th signal can generate the spatio-temporal equivalent virtual array signal through the above derivation.

[0025] Considering a uniform linear array of M elements with a spacing of d, when P signals are incident on the virtual array, the overall received signal X(t) = [x1(t), x2(t),..., x M (t)] T = A′(θ)S′(t) + N′(t), where: x i (t) is the i-th received signal, A(θ) = [a(θ1), a(θ2),..., a(θ P )] is the array manifold matrix, is the steering vector corresponding to the angle θ i , S(t) = [s1(t), s2(t), …, s P (t)] T is the signal source matrix, and N(t) represents white noise.

[0026] Since the overall signal χ(t) and X(t) are highly similar, and the different frequency components of the human chest surface motion can be equivalent to the different arrival direction angles of the electromagnetic waves incident on the virtual array, the equivalent condition between the two is θ i = sin(2Kf i / f s ) -1 , where: θ i is the i-th angle information of the incoming wave angle, f i is the frequency value of the i-th component in the human chest cavity phase information, f s is the radar signal sampling rate.

[0027] Since the sin(·) function has greater sensitivity near the 0-degree angle, the larger the value of K, the more accurate the estimated angle value. To avoid angle ambiguity, there is an upper limit for the value of K, that is, ω i Kt s ≤π. From this, it can be summarized that the time-space virtual array equivalence principle includes: the equivalence between the incoming wave angle and the frequency of the chest movement signal: θ i = sin(2Kf i / f s ) -1 , and the value of the optimal virtual array length K is where f Max is the frequency value of the component with the largest frequency in the chest displacement signal d(t).

[0028] As Figure 4 shown, the frequency estimation method from coarse to fine specifically includes the following steps:

[0029] Step 1: Obtain the single-channel phase information containing displacement.

[0030] Step 2: Let K = 1, that is, no data segmentation and recombination is performed. At this time, the number of corresponding virtual arrays is equal to the number of the input single-channel phase information, that is, the number of virtual arrays is the largest at this time.

[0031] Step 3: Use the data with the largest number of virtual arrays to perform a rough estimation of the heart rate signal. Specifically, use the MUSIC method to perform spectral estimation on the virtual array data. Since the signal length in each virtual array is 1, the signal-to-noise ratio obtained by the MUSIC method is relatively poor, so it is called a rough estimation. According to the spectral result of the rough estimation, the approximate interval of the target heart rate value can be found. Specifically, find the maximum value point of the spectral peak in the interval [0.1Hz, 0.3Hz] and consider it as the rough estimated heart rate value f c .

[0032] Step 4: Update according to the heart rate value f c obtained by the rough estimation, where f s is the sampling frequency of the radar.

[0033] Step 5: According to the new K value, perform data segmentation and recombination to obtain a new virtual array signal matrix. Then use the ESPRIT method to estimate the arrival angles of the waves for the virtual array signal matrix. Since in the ESPRIT method, the number of arriving waves needs to be more than the vital sign parameters to be estimated, namely the respiration rate and heart rate, it is assumed that the number of arriving waves is 5. Obtain the first five arrival angles using the ESPRIT method, Take the first f within the interval [0.1Hz, 0.3Hz] i as the value of the respiration rate, and the first f within the interval [0.3Hz, 3Hz] i as the value of the heart rate, and calculate the heart rate value R r = f i * 60.

[0034] Step 6: When the average estimated value of the heart rate per minute is R H . Judge that if |R r - R H |≥ε, it means that the heart rate has changed rapidly and the value of K needs to be updated again, that is, repeat Step 2. Otherwise, the heart rate has not changed rapidly and the current K value is still maintained for the next heart rate estimation, so repeat Step 4.

[0035] After specific actual experiments, in the laboratory environment, the experimental settings are as Figure 5 shown. The subject sits with the back against the backrest and remains stationary, breathing normally. The radar is fixed 1m away from the subject through a floor stand, and the radar is directed directly at the subject's chest. A finger clip PPG is worn on the subject's right index finger as the gold standard reference for the heart rate. The PPG signal is directly connected to the computer, and the radar data is synchronously collected with the computer, and the sampling rate is 100Hz for both. The parameters of the radar system used are: the operating frequency is 24GHz, single-frequency continuous wave transmission, single transmit and single receive, and the EIRP is 13dBm. The experimental data obtained are:

[0036] As Figure 6 shown, it is a comparison chart of the spectral results of this embodiment and the existing frequency-based heart rate estimation algorithms. In the figure, the PPG signal represents the gold standard. The compared frequency-based heart rate estimation algorithms are the DFT method, the roughly estimated MUSIC-Ori method, and the coarse-to-fine spatio-temporal equivalent virtual array method proposed by the present invention. The respiration rate obtained from the figure is 0.253Hz, and the heart rate is 1.261Hz. The heart rate obtained by the DFT method is 1.35Hz, and the heart rate frequency cannot be accurately estimated. The roughly estimated MUSIC-Ori method cannot form a peak at the heart rate frequency, while the method proposed by the present invention can accurately estimate the heart rate frequency as 1.261Hz, which is consistent with the gold standard.

[0037] As Figure 7As shown, it is a comparison chart of the 10-minute heart rate estimation results of this embodiment and the existing frequency-based heart rate estimation algorithm. Due to resolution limitations, the estimation result of the DFT method is the worst, with RMSE = 5.72. The result obtained by the MUSIC-Ori algorithm achieved better accuracy in some time periods, but had large fluctuations, with an overall RMSE = 2.31. The proposed method based on the spatio-temporal equivalent virtual array technology performed the best, with high accuracy and high stability throughout the time period, RMSE = 0.875. Compared with the traditional DFT method, the estimation accuracy was improved by more than 12%.

[0038] Compared with the prior art, since the subspace method in DOA estimation is adopted for frequency estimation in the present invention, it can obtain super-resolution heart rate estimation results, and thus can detect vital sign signals more accurately. At the same time, the estimation process from coarse to fine makes the overall heart rate estimation process more robust, greatly reducing the probability of false estimation and missed estimation. When using this example for implementation, for the 10-minute sitting heart rate estimation of more than 10 people, the RMSE of the average error of the heart rate estimation obtained by this method compared with the gold standard PPG result is only 0.81 beats per minute.

[0039] The above specific implementation can be locally adjusted in different ways by those skilled in the art without departing from the principles and purposes of the present invention. The protection scope of the present invention is subject to the claims and is not limited by the above specific implementation. All implementation solutions within its scope are subject to the present invention.

Claims

1. A vital sign detection method based on spatio-temporal equivalent virtual array technology, characterized in that, Based on the phase information demodulated from the received single-channel continuous-wave radar signal, an equivalent multi-channel virtual array signal is obtained through signal recombination; then, the subspace method is used to perform frequency-domain estimation on the multi-channel virtual array signal to obtain the angle-of-arrival information of the incoming wave; according to the target frequency converted from the angle-of-arrival information, the vital sign information of the target is accurately obtained through a coarse-to-fine frequency estimation method; The signal recombination mentioned above refers to: cutting the phase information of a single channel in the time dimension. Specifically: the phase information matrix at is decomposed into sub-signals with a length of . Among them, the th sub-signal is , is the sampling time interval, is the sampling rate, and each sub-signal is an equivalent virtual array, that is, an element of an equivalent SIMO radar; The coarse-to-fine frequency estimation method specifically includes: Step 1: Obtain the phase information of a single channel containing displacement; Step 2. Let , that is, no data splitting and recombination is performed. At this time, the number of corresponding virtual arrays is equal to the number of single-channel phase information inputs, that is, the number of virtual arrays is the largest at this time; Step 3: Coarsely estimate the heart rate signal using the data of the maximum number of virtual arrays. Specifically, use the MUSIC method to perform spectral estimation on the virtual array data. Since the signal length in each virtual array is 1, the signal-to-noise ratio obtained by the MUSIC method is poor, so it is called a coarse estimate. According to the spectral result of the coarse estimate, find the maximum value point of the spectral peak in the heart rate value interval [0.1 Hz, 0.3 Hz] of the target, and consider it as the coarsely estimated heart rate value ; Step 4, according to the heart rate value obtained by rough estimation , update ; where is the sampling frequency of the radar; Step 5: According to the new K value, perform data segmentation and recombination to obtain a new virtual array signal matrix; and use the ESPRIT method to estimate the arrival angle of the virtual array signal matrix. Since in the ESPRIT method, the number of incoming waves needs to be more than the vital sign parameters to be estimated: respiratory rate and heart rate, it is assumed that the number of incoming waves is 5. Use the first five arrival angles obtained by the ESPRIT method, , take the first one in the interval [0.1Hz, 0.3Hz] as the value of the respiratory rate, and the first one in the interval [0.3Hz, 3Hz] as the value of the heart rate, and calculate the heart rate value ; Step 6. When the average estimated value of the heart rate per minute is ; judge that if , the heart rate has changed rapidly and the value of K needs to be updated again, that is, repeat Step 2; otherwise, the heart rate has not changed rapidly, and the current value of K is still maintained for the next heart rate estimation, so repeat Step 4.

2. The vital sign detection method based on the spatio-temporal equivalent virtual array technology according to claim 1, wherein The demodulation refers to: for the received single-channel continuous-wave radar signal, first mix and down-convert the received signal with the transmitted signal to obtain two orthogonal intermediate-frequency signals, then perform circular fitting on the I and Q signals to complete amplitude and offset calibration; then use the MDACM algorithm for phase demodulation to obtain the phase information of the target.

3. The vital sign detection method based on the spatio-temporal equivalent virtual array technology according to claim 1, characterized in that, The subspace method refers to: using the subspace estimation (MUSIC) algorithm and the ESPRIT algorithm in the theory of angle-of-arrival estimation to estimate the received signal matrix composed of M array elements of the SIMO radar to obtain the angle-of-arrival information.

4. The vital sign detection method based on the spatio-temporal equivalent virtual array technology according to claim 1, wherein, The conversion mentioned above means that the angle information of the incoming wave obtained by the subspace method is equivalent to the frequency values of the respective frequency components in the phase signal of the human chest cavity, specifically: , where: is the i-th angle information of the incoming wave angle, is the frequency value of the i-th component in the human chest cavity phase information, is the radar signal sampling rate; When the received single-channel radar phase signal , whose magnitude is , after its segmentation, sub-signals are generated, where the magnitude of each sub-signal is . The -th sub-signal matrix can . Thus, the conversion of the single-channel SISO radar signal to the multi-channel SIMO radar virtual channel signal is completed.

5. The vital sign detection method based on the spatio-temporal equivalent virtual array technology according to claim 1, characterized in that The frequency estimation method from coarse to fine refers to: when reconstructing the signal, select a smaller length of the sub-signal according to the MUSIC method , according to the heart rate value obtained by the coarse estimation, use the ESPRIT method to optimize the length of the sub-signal , until the length of each sub-signal corresponding to the frequency component to be estimated currently is the best selection value, and accurate heart rate and respiratory rate estimation results are achieved.

Images