Millimeter wave radar vital sign detection system based on SVMD and harmonic estimation

Abstract

The invention relates to the technical field of non-contact vital sign detection, and discloses a millimeter-wave radar vital sign detection system based on SVMD and harmonic estimation, which comprises the following steps: placing a millimeter-wave radar at a position right ahead of a detected human body to collect an IF signal, obtaining a corresponding distance between the millimeter-wave radar and the detected human body in a field of view according to the IF signal, and calculating the distance between the millimeter-wave radar and the detected human body according to the distance; determining the position of a target human body in the area; selecting a candidate distance unit adjacent to the target human body according to the phase information of the IF signal, performing principal component analysis by combining the distance unit where the target human body is located and the phase information of the IF signal in the candidate distance unit, and extracting a vital sign signal; and decomposing the vital sign signal and reconstructing to obtain a respiration signal, a heartbeat signal and a heartbeat harmonic signal, and performing frequency spectrum estimation and correction on the reconstructed signals to obtain a final respiration estimation value and a final heartbeat estimation value. According to the invention, vital sign signals can be effectively enhanced and extracted, and accurate estimation of respiration and heartbeat is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of non-contact vital sign detection, and in particular to a millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation. Background Art

[0002] Human vital signal detection is a technology that monitors and processes key physiological indicators of the human body, including heartbeat, respiration, body temperature, blood pressure, etc., which together constitute the vital signals of the human body. This detection technology has a wide range of applications in the fields of medical care, health management, sports training, and security. In recent years, the detection of respiration and heartbeat signals has gradually developed from contact detection to non-contact detection, which can achieve non-invasive, continuous, and real-time vital sign monitoring, greatly improving the efficiency and accuracy of detection. At present, vital sign detection based on millimeter wave radar has become a major research direction. Millimeter wave radar has excellent anti-electronic interference performance, high speed measurement accuracy and resolution, and can be used in scenarios that require accurate measurement of speed and distance. It also has all-weather and all-day working characteristics and relatively low cost. The technology based on millimeter wave radar makes up for the shortcomings of traditional optical cameras that are easily affected by light and obstructed by objects, and pays more attention to privacy protection; at the same time, it also makes up for the shortcomings of traditional ultrasonic radars that cannot penetrate non-metallic objects, and is suitable for scenarios where vital sign signal detection of the human body is performed.

[0003] The human body can be used as a target radar monitoring sensor. The skin expansion movement caused by human breathing / heartbeat will cause the radar echo parameters to change, such as the common fluctuations based on the instantaneous phase of the echo signal and the inversion of breathing / heart rate parameters. FMCW (Frequency Modulated Continuous Wave Radar) radar sends multiple linear frequency modulated pulses to the human body according to a predetermined time interval. Each pulse performs a distance fast Fourier transform and selects a distance unit corresponding to the position of the human chest; each linear frequency modulated pulse records the signal phase in the selected distance unit, thereby calculating the phase change and deriving the speed; the remaining signals contain vital signs signals, and the breathing and heartbeat signals can be obtained separately through algorithm separation and extraction.

[0004] In the prior art, there are methods based on radar technology to separate human breathing and heartbeat parameters, mainly including EWT (Empirical Wavelet Transform), VMD (variational mode decomposition), machine learning-based methods, neural network-based methods, etc. However, these methods all have problems such as incomplete noise signal filtering, incomplete vital sign signal extraction, and inadequate processing of non-stationary signals, which have a great impact on the quality of the obtained signal and lead to misestimation of breathing and heartbeat. Summary of the invention

[0005] To this end, the technical problem to be solved by the present invention is to overcome the deficiencies in the prior art and provide a millimeter-wave radar vital signs detection system based on SVMD and harmonic wave estimation, which can effectively enhance and extract vital signs signals and achieve accurate estimation of breathing and heartbeat.

[0006] In order to solve the above technical problems, the present invention provides a millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation, comprising:

[0007] The millimeter wave radar is placed in front of the human body to be detected and is used to collect the vital signs of the human body to obtain the IF signal;

[0008] A target determination module is used to obtain the distance between the millimeter wave radar and the human body detected in the field of view according to the IF signal, and determine the position of the target human body in the area through a target detection algorithm;

[0009] A candidate distance unit selection module is used to select a candidate distance unit adjacent to the target human body according to the phase information of the IF signal;

[0010] A vital sign signal extraction module is used to perform principal component analysis on the phase information of the IF signal in the distance unit where the target human body is located and the phase information of the IF signal in the candidate distance unit, and extract the component with the largest variance as the vital sign signal;

[0011] A signal reconstruction module, used for decomposing the vital sign signal and reconstructing a breathing signal, a heartbeat signal and a heartbeat harmonic signal;

[0012] The estimation module is used to perform spectrum estimation and correction on the reconstructed breathing signal, heartbeat signal and heartbeat harmonic signal to obtain the final breathing estimation value and the final heartbeat estimation value.

[0013] Further, selecting a candidate distance unit adjacent to the target human body according to the phase information of the IF signal includes:

[0014] The overlapping sliding window algorithm is used to extract the phase information of the IF signal in the distance unit where the target human body is located and the phase information of the IF signal in the adjacent distance unit. The cross-correlation of the phase information of the IF signal in the distance unit where the target human body is located and the phase information of the IF signal in the adjacent distance unit is calculated, and the adjacent distance unit whose cross-correlation result is greater than the preset threshold is selected as the candidate distance unit.

[0015] Further, decomposing the vital sign signal includes:

[0016] The vital sign signal is divided into a respiratory frequency component, a heartbeat frequency component and a noise frequency component, and the component that conforms to the respiratory frequency component is selected as the first mode, the classification that conforms to the heartbeat frequency component is selected as the second mode, and the component in high-quality harmonics is selected as the noise frequency component;

[0017] The SVMD algorithm is used to extract the modes one by one until the reconstruction error reaches a certain threshold, and the target modal components of the respiratory frequency component and the heart rate frequency component are obtained.

[0018] Furthermore, when the SVMD algorithm is used to extract the modes one by one,

[0019] The first part of the noise frequency component is 0, and the update method of the mode in the n+1th iteration is:

[0020]

[0021] in, is the estimated value of the Lth mode obtained in the n+1th iteration, L represents the number of decomposed modes, n represents the number of iterations, ω represents the current frequency, is the estimated value of the Lth mode obtained at the nth iteration, is the mapping of the original input signal in the frequency domain; is the center frequency of the Lth mode obtained in the nth iteration, ω i is the center frequency of the ith mode; is the Lagrange multiplier corresponding to ω, α represents the balance parameter of the data fidelity constraint, and α can be obtained by the Lagrange multiplier method.

[0022] Furthermore, the iterative update method of the center frequency of the Lth mode is:

[0023]

[0024] in, represents the center frequency of the Lth mode obtained in the n+1th iteration, is the estimated value of the Lth mode obtained at the n+1th iteration.

[0025] Furthermore, the iterative update method of the Lagrange multiplier is:

[0026]

[0027] Among them, λ n+1 is the Lagrange multiplier obtained at the n+1th iteration, λ n is the Lagrange multiplier obtained in the nth iteration, τ represents the update parameter, is the estimated value of the ith mode obtained at the nth iteration, is the estimated value of the ith mode, is the value of the ith mode determined in the n+1th iteration, is the center frequency of the ith mode obtained in the n+1th iteration.

[0028] Furthermore, the reconstruction obtains a respiratory signal, a heartbeat signal and a heartbeat harmonic signal. Specifically, after selecting the optimal parameter group of the target modal components of the respiratory frequency component and the heartbeat frequency component, the vital sign signal is modally decomposed to obtain an inherent modal function, and the reconstructed respiratory signal, the reconstructed heartbeat fundamental frequency signal, the reconstructed heartbeat second harmonic signal and the reconstructed heartbeat third harmonic signal are obtained according to the inherent modal function.

[0029] Furthermore, the reconstruction process of the reconstructed respiratory signal is specifically as follows:

[0030] Perform FFT analysis on each intrinsic mode function to obtain the total energy of each intrinsic mode function component;

[0031] Calculate the total energy of each intrinsic mode function component and the total energy within the respiratory frequency band, and construct the judgment condition as follows:

[0032]

[0033] Where E(i) is the total energy of each intrinsic mode function component, E r (i) is the total energy within the respiratory frequency band, is an empirical parameter; if the judgment condition is met, the intrinsic modal function component is considered to be an effective component of the respiratory signal, and the intrinsic modal function components c of all respiratory signals that meet the judgment condition are selected 1 (t),c 2 (t),…,c j (t),…,c k (y), c j (t) is the intrinsic modal function component of the jth respiratory signal that meets the judgment condition at time t, and k is the number of intrinsic modal function components of the respiratory signal that meets the judgment condition;

[0034] The reconstructed respiratory signal is calculated as:

[0035] d r =d r (1),d r (2),...,d r (t),...,d r (N),

[0036]

[0037] Among them, d r is the reconstructed respiratory signal, d r (t) is the reconstructed respiratory signal at time t, and N is the total time.

[0038] Furthermore, the final breathing estimation value is obtained by:

[0039] The reconstructed breathing signal is subjected to fast Fourier transform to convert the time domain signal into a frequency domain signal, and the peak value of the frequency domain signal is selected as the final breathing estimation value.

[0040] Furthermore, the final heartbeat estimation value is obtained by:

[0041] Performing fast Fourier transform on the reconstructed heartbeat fundamental frequency signal, the reconstructed heartbeat second harmonic signal, and the reconstructed heartbeat third harmonic signal to convert the time domain signal into a frequency domain signal, and selecting the peak value of the frequency domain signal as the corresponding heartbeat fundamental frequency estimation frequency, heartbeat second harmonic estimation frequency, and heartbeat third harmonic estimation frequency;

[0042] The estimated frequency of the heartbeat fundamental frequency, the estimated frequency of the heartbeat second harmonic, and the estimated frequency of the heartbeat third harmonic are used as the input of the Kalman filter, and the measurement vector is defined as:

[0043]

[0044] Among them, z h is the measurement vector, f h1 is the estimated frequency of the heartbeat fundamental frequency, f h2 is the estimated frequency of the second harmonic of the heartbeat, f h3 is the estimated frequency of the third harmonic of the heartbeat, and T represents the transposition operation;

[0045] The state vector is constructed as:

[0046] x=[f h f h ′ f h ″] T ,

[0047] Among them, x is the state vector, f h =f h1 , f h ′ is f h1The first derivative of h ″ is f h1 The second derivative of

[0048] According to the current state of the mth sliding window, the predicted state is calculated using the state transfer equation:

[0049]

[0050] in, is the current state of the mth sliding window, is the predicted state in the m+1th sliding window, F is the state transfer matrix, and Δt is the duration of the non-overlapping part of the sliding window;

[0051] Denote the state vector as having a mean The covariance is P m,m The Gaussian process of , the predicted covariance is:

[0052] P m+1,m =FP m,m F T +Q,

[0053] Among them, P m+1,m is the predicted covariance, Q is the covariance of the system process noise, is the process noise;

[0054] Construct the observation matrix, denoted as H,

[0055] Construct the measurement error, denoted as e m ,

[0056] Construct the relevant covariance, denoted as S m , S m =HP m+1,m H T +R; where R is a diagonal matrix containing the uncertainty of each heartbeat estimate region, is the expected square of the estimated error of the heartbeat fundamental frequency, is the expected square of the estimated error of the estimated frequency of the second harmonic of the heartbeat, is the expected square of the estimated error of the estimated frequency of the third harmonic of the heartbeat;

[0057] Initialize the predefined Q and R, the initial state is The covariance is P 0,0 If there is a heartbeat estimate close to the predicted state Then in z h Select the state close to the prediction The heartbeat estimate is taken as the final heartbeat estimate; the distance between the new measurement vector and the filter prediction state is minimized, and the state vector is updated to z is the index corresponding to the selected heartbeat estimate in the measurement vector, e m (z) represents e m The zth element in k m is the Kalman gain, h z is the transpose of the zth row of H, S m (z,z) is S m The z-th diagonal element of ; the relevant covariance update for the next filter iteration, denoted as P m+1,m+1 , I is the identity matrix.

[0058] The above technical solution of the present invention has the following beneficial effects compared with the prior art:

[0059] The present invention uses cross-correlation and principal component analysis methods to effectively utilize human vital sign signals dispersed in different distance units to enhance respiratory and heartbeat signals; on this basis, the vital sign signals are decomposed and reconstructed, and respiratory and heartbeat are estimated based on the reconstructed signals, thereby achieving effective separation and accurate estimation of respiratory and heartbeat signals. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to make the content of the present invention more clearly understood, the present invention is further described in detail below according to specific embodiments of the present invention in conjunction with the accompanying drawings, wherein:

[0061] Figure 1 This is a detection process diagram of a millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation in a preferred embodiment of the present invention.

[0062] Figure 2 The figure shows the time domain waveform comparison of the original phase signal and the signal enhanced by cross-correlation and PCA.

[0063] Figure 3 The spectrum comparison diagram of the original phase signal and the signal enhanced by cross-correlation and PCA.

[0064] Figure 4 It is a comparison curve diagram of the final breathing estimation value and the actual breathing monitoring value obtained through the simulation experiment in the embodiment of the present invention.

[0065] Figure 5 It is a comparison curve diagram of the final heartbeat estimation value and the actual heartbeat monitoring value obtained through the simulation experiment in the embodiment of the present invention. DETAILED DESCRIPTION

[0066] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it, but the embodiments are not intended to limit the present invention.

[0067] Embodiment 1

[0068] Reference Figure 1 As shown, the present invention discloses a millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation, including a millimeter wave radar, a target determination module, a candidate distance unit selection module, a vital sign signal extraction module, a signal reconstruction module and an estimation module. The process of vital sign detection is:

[0069] S1: Place the millimeter-wave radar in front of the human body to be detected, and collect the vital signs of the human body to obtain the IF signal. The RF transceiver module uses a millimeter-wave radar above 60Ghz to transmit microwave signals to the living body and receive the echo signal of the living body within the detection range. After passing through the coupler, the signal is transmitted to the human body by the transmitting antenna, and then the echo signal is received by the receiving antenna. After amplification, the echo signal is mixed with the transmitting signal in the mixer to obtain an intermediate frequency signal, which is then transmitted to the signal processing module through the serial port after analog-to-digital conversion; the signal processing device is used to process the received original echo data to form the life parameter data of the living body.

[0070] S2: The target determination module obtains the distance between the millimeter wave radar and the human body detected in the field of view according to the IF signal, and determines the position of the target human body in the area through the target detection algorithm.

[0071] The IF signal is a radar intermediate frequency signal after mixing and sampling. The IF signal is mean filtered to remove the DC bias caused by the millimeter wave radar hardware itself and the static clutter interference in the environment. Then the mean filtered signal is fast Fourier transformed, and a range-time map (RTM) is constructed according to the distance distribution. The position of the target human body in the area is determined by the CA-CFAR algorithm in combination with the range-time map.

[0072] S3: After determining the existence of the target, the candidate distance unit selection module selects the candidate distance unit adjacent to the target human body according to the phase information of the IF signal.

[0073] The overlapping sliding window algorithm is used to extract the phase information of the IF signal in the distance unit where the target human body is located and the phase information of the IF signal in the adjacent distance unit. The cross-correlation of the phase information of the IF signal in the distance unit where the target human body is located and the phase information of the IF signal in the adjacent distance unit is calculated, and the adjacent distance unit whose cross-correlation result is greater than the preset threshold is selected as the candidate distance unit.

[0074] In this embodiment, the Pearson cross-correlation coefficient is used to calculate the cross-correlation coefficient between the phase information of the IF signal in the distance unit where the target human body is located and the phase information of the IF signal in the adjacent distance unit:

[0075]

[0076] Where ρ is the mutual correlation coefficient, It is the phase information of the IF signal in the distance unit where the target human body is located. is the phase information of the IF signal in the adjacent distance unit, and E() represents the expected value of the corresponding variable. The adjacent distance unit with a correlation coefficient greater than a threshold is selected as a candidate distance unit, and the preset threshold in the present invention is selected as 0.75.

[0077] S4: The vital sign signal extraction module performs principal component analysis on the phase information of the IF signal in the distance unit where the target human body is located and the phase information of the IF signal in the candidate distance unit, and extracts the component with the largest variance as the vital sign signal.

[0078] The phase information of the IF signal in the candidate distance unit is subjected to principal component analysis to extract the vital sign signals in different distance units and enhance the vital sign signals. The principal component analysis method used in this embodiment is the PCA algorithm. PCA is one of the most classic linear subspace dimensionality reduction methods. It mainly removes the correlation of data through linear transformation and obtains low-dimensional data. The optimization goal is to maximize the variance after projection. The specific process is:

[0079]

[0080] Among them, P is the extracted principal component, is the phase information of the IF signal in all candidate distance units, for The covariance of for The principal component with the largest variance is selected as the enhanced vital sign signal.

[0081] The eigenvector corresponding to the maximum eigenvalue is selected as the enhanced vital sign signal for the next step of processing. The PCA global variance maximization objective function considers the linear space projection of the data and enhances the global distribution characteristics by intercepting an appropriate number of eigenvectors (retaining most of the information of the original data). Figure 2 This is a time domain waveform comparison diagram of the original phase signal and the signal enhanced by cross-correlation and PCA. Figure 3 The spectrum comparison diagram of the original phase signal and the signal after cross-correlation and PCA enhancement shows that the vital sign signal has been enhanced.

[0082] S5: The signal reconstruction module decomposes the vital sign signal to reconstruct a respiratory signal, a heartbeat signal and a heartbeat harmonic signal. In this embodiment, the continuous variational mode decomposition (SVMD) algorithm is used to decompose and reconstruct the vital sign signal to obtain the respiratory signal, the heartbeat signal and the heartbeat harmonic signal.

[0083] S5-1: The vital sign signal contains multiple modal components. In this embodiment, the vital sign signal is divided into three frequency components, namely, the respiratory frequency component, the heartbeat frequency component and the noise frequency component. The component that meets the respiratory frequency component is selected as the first modality, the component that meets the classification of the heartbeat frequency component is selected as the second modality, and the component in high-quality harmonics is selected as the noise frequency component; the last three parts of the signal are the sum of two constant frequency conversions and a cosine wave of a chirp signal, that is, the input signal f(t) is decomposed into three parts: the respiratory frequency component u(t), the heartbeat frequency component g(t) and the noise frequency component (or residual signal) f r (t), its expression is:

[0084] f(t)=u(t)+g(t)+f r (t),

[0085] Among them, the residual signal f r (t) is the input signal other than breathing and heartbeat. u(t) and g(t) each contain two parts: u(t) includes the incompletely decomposed u(t) (denoted as u i (t)) and f r The unprocessed respiratory-related signal in (t), g(t) includes g(t) when it is not completely decomposed (denoted as g i (t)) and f r Unprocessed heartbeat-related signal in (t).

[0086] S5-2: Use the SVMD algorithm to extract modes one by one until the reconstruction error reaches a certain threshold, and obtain the target modal components of the respiratory frequency component and the heart rate frequency component, which has lower computational complexity.

[0087] In order to find the first mode of the respiratory frequency component, let the noise frequency component f r The first part of (t) is 0, and the update method of the mode in the n+1th iteration of SVMD is:

[0088]

[0089] in, is the estimated value of the Lth mode obtained in the n+1th iteration, L represents the number of decomposed modes, n represents the number of iterations, ω represents the current frequency, is the estimated value of the Lth mode obtained at the nth iteration, is the mapping of the original input signal f(t) in the frequency domain; is the center frequency of the Lth mode obtained in the nth iteration. Similarly, ω i is the center frequency of the ith mode; is the Lagrange multiplier corresponding to ω, α represents the balance parameter of the data fidelity constraint, and α can be obtained by the Lagrange multiplier method;

[0090] The iterative update method of the center frequency of the Lth mode can be approximately expressed as:

[0091]

[0092] in, represents the center frequency of the Lth mode obtained in the n+1th iteration, is the estimated value of the Lth mode obtained at the n+1th iteration.

[0093] Finally, the iterative update of the Lagrange multiplier λ is achieved by the double ascent method, which is calculated as:

[0094]

[0095] Among them, λ n+1 is the Lagrange multiplier obtained at the n+1th iteration, λ n is the Lagrange multiplier obtained in the nth iteration, τ represents the update parameter, is the estimated value of the ith mode obtained at the nth iteration, is the estimated value of the ith mode, is the value of the ith mode determined in the n+1th iteration, is the center frequency of the ith mode obtained in the n+1th iteration.

[0096] The SVMD algorithm extracts modal components one by one until the reconstruction error reaches a threshold, and the result is the target modal component. The SVMD algorithm will extract relevant human signals such as heartbeat and breathing by continuously applying the variational modal extraction method to the dynamic signals of human vital signs. By adding constraints to avoid convergence to the previously extracted mode, the algorithm continues to extract the original signal until all modes are extracted or the reconstruction error (the error between the input signal and the sum of the modes) is less than a threshold.

[0097] SVMD uses an iterative method to decompose the signal into a set of intrinsic mode functions (IMFs) to represent the different frequency components or modes of the signal. IMFs can be obtained by solving an optimization problem that minimizes the cross-interference between modes.

[0098] The SVMD algorithm starts by initializing the decomposition process with a set of initial IMFs, and then iterates by alternating between updating the IMFs and estimating the residual signal until convergence is reached. The updated IMFs capture the target pattern of the signal, while the residual represents the signal components that cannot be further decomposed.

[0099] The advantage of SVMD over traditional decomposition methods is that it can handle non-stationary signals with time-varying frequency components. It can effectively separate signals with complex spectral content or transient behavior. Compared with existing algorithms, a major advantage of SVMD over VMD is that it does not need to determine the number of modes of the vital sign signal to obtain relatively accurate respiratory and heartbeat separation signals. The correlation coefficient (CC) between the reconstructed signal (the sum of the extracted modes) and the original noise-free signal is an important indicator for measuring the accuracy of the reconstructed signal. As a method that does not require knowing the number of modes, the CC value that SVMD can obtain is comparable to the highest value that VMD can obtain. In addition, compared with traditional VMD, another advantage of SVMD in separating the target heartbeat and respiratory signals is that the computational complexity is much lower than VMD (about 10 times on average), and the number of iterations of the converged mode is less than VMD, which ensures the real-time performance of the signal.

[0100] Different from VMD, in SVMD, the value of the total number of modes K and the weight factor α do not need to be determined and set, and only a minimum allowable value and a maximum allowable value need to be set for α. In the present invention, the minimum value of α is set to 100 and the maximum value is set to 4000.

[0101] S5-3: Reconstruct the respiratory signal, heartbeat signal and heartbeat harmonic signal. Specifically, after selecting the optimal parameter group of the target modal components of the respiratory frequency component and the heartbeat frequency component, perform modal decomposition on the vital sign signal to obtain the intrinsic modal function (IMF), and obtain the reconstructed respiratory signal, reconstructed heartbeat fundamental frequency signal, reconstructed heartbeat second harmonic signal and reconstructed heartbeat third harmonic signal according to the intrinsic modal function.

[0102] The reconstruction process of the reconstructed respiratory signal is as follows:

[0103] 1) Perform FFT (fast Fourier transform) analysis on each intrinsic mode function (IMF) to obtain the total energy of each intrinsic mode function component.

[0104] 2) Calculate the total energy of each intrinsic mode function component and the total energy within the respiratory frequency band, and construct the judgment condition as follows:

[0105]

[0106] Where E(i) is the total energy of each intrinsic mode function component, E r (i) is the total energy within the respiratory frequency band, is an empirical parameter; if If the value is too large, the useful IMF components will be discarded. If the value is too small, useless IMF components will be introduced. The best effect is achieved when the value is 0.6. If the judgment condition is met, the intrinsic modal function component is considered to be an effective component of the respiratory signal. The intrinsic modal function components c of all respiratory signals that meet the judgment condition are selected. 1 (t),c 2 (t),…,c j (t),…,c k (y), c j (t) is the intrinsic modal function component of the jth respiratory signal that meets the judgment condition at time t, and k is the number of intrinsic modal function components of the respiratory signal that meets the judgment condition.

[0107] 3) Add up all the effective components of the respiratory signal and calculate the reconstructed respiratory signal as:

[0108] d r =d r (1),d r (2),...,d r (t),...,d r (N),

[0109]

[0110] Among them, d r is the reconstructed respiratory signal, d r (t) is the reconstructed respiratory signal at time t, and N is the total time.

[0111] The reconstructed heartbeat fundamental frequency signal is denoted as d h1 , the reconstructed heartbeat second harmonic signal, denoted as d h2 , the reconstructed heartbeat third harmonic signal, denoted as d h3The reconstruction method is the same as the reconstructed respiratory signal. Considering the breathing and heartbeat frequency of the human body under normal conditions, 2) the respiratory frequency band range is 0.1-0.6Hz, the heartbeat fundamental frequency signal band range is 0.8-1.5Hz, the heartbeat second harmonic frequency band range is 1.6-3Hz, and the heartbeat third harmonic frequency band range is 2.4-4.5Hz.

[0112] S6: Estimation module, used for performing spectrum estimation and correction on the reconstructed breathing signal, heartbeat signal and heartbeat harmonic signal to obtain the final breathing estimation value and the final heartbeat estimation value. In this embodiment, the spectrum of the breathing signal, heartbeat signal and heartbeat harmonic signal is estimated to obtain the breathing rate, heart rate and heartbeat harmonic frequency; the heart rate and heartbeat harmonic frequency are input into the Kalman filter for correction to obtain the final accurate heart rate estimation value.

[0113] The final breathing estimate is obtained by: r Perform fast Fourier transform to convert the time domain signal into frequency domain signal, and select the peak value of the frequency domain signal as the final breathing estimation value, denoted as f r .

[0114] The final heart rate estimate is obtained as follows:

[0115] 1) Reconstruct the heartbeat baseband signal d h1 , reconstructed heartbeat second harmonic signal d h2 , reconstructed heartbeat third harmonic signal d h3 Perform fast Fourier transform to convert the time domain signal into frequency domain signal, and select the peak value of the frequency domain signal as the corresponding heartbeat fundamental frequency estimation frequency (denoted as f h1 ), the estimated frequency of the second harmonic of the heartbeat (denoted as f h2 ), the estimated frequency of the third harmonic of the heartbeat (denoted as f h3 );

[0116] 2) Since there are multiple heart rate estimation results, it is necessary to select them. The heart rate fundamental frequency estimation frequency, heart rate second harmonic estimation frequency, and heart rate third harmonic estimation frequency are used as the input of the Kalman filter, and the measurement vector is defined as:

[0117]

[0118] Among them, z h is the measurement vector, f h1 is the estimated frequency of the heartbeat fundamental frequency, f h2 is the estimated frequency of the second harmonic of the heartbeat, f h3 is the estimated frequency of the third harmonic of the heartbeat, and T represents the transposition operation;

[0119] The state vector is constructed as:

[0120] x=[f h ′ f h ′ f h ″] T ,

[0121] Among them, x is the state vector, f h =fh 1 , f h ′ is fh 1 The first derivative of h ″ is f h1 The second derivative of

[0122] According to the current state of the mth sliding window, the predicted state is calculated using the state transfer equation:

[0123]

[0124] in, is the current state of the mth sliding window, is the predicted state in the m+1th sliding window, F is the state transfer matrix, and Δt is the duration of the non-overlapping part of the sliding window;

[0125] Denote the state vector as having a mean The covariance is P m,m The Gaussian process of , the predicted covariance is:

[0126] P m+1,m =FP m,m F T +Q,

[0127] Among them, P m+1,m is the predicted covariance, Q is the covariance of the system process noise, and is related to the uncertainty of the model. is the process noise;

[0128] Construct the observation matrix, denoted as H,

[0129] Construct the measurement error, denoted as e m ,

[0130] Construct the relevant covariance, denoted as S m , S m =HP m+1,m H T +R; where R is a diagonal matrix containing the uncertainty of each heartbeat estimate region, is the expected square of the estimated error of the heartbeat fundamental frequency, is the expected square of the estimated error of the estimated frequency of the second harmonic of the heartbeat, is the expected square of the estimated error of the estimated frequency of the third harmonic of the heartbeat;

[0131] 3) Initialize the predefined Q and R, the initial state is The covariance is P 0,0 If there is a heartbeat estimate close to the predicted state Then in the measurement vector z h Select the state close to the prediction The heartbeat estimate is taken as the final heartbeat estimate; the distance between the new measurement vector and the filter prediction state is minimized, and the state vector is updated to z is the index corresponding to the selected heartbeat estimate in the measurement vector, e m (z) represents e m The zth element in k m is the Kalman gain, h z is the transpose of the zth row of H, S m (z,z) is S m The z-th diagonal element of ; the relevant covariance update for the next filter iteration, denoted as P m+1,m+1 , I is the identity matrix.

[0132] In this embodiment, the initialization parameter of the Kalman filter is defined as ρ a =2,σ r1 =σ r2 =σ r3 =1.5, P 0,0 =1000I. Considering that the pulse respiratory quotient of the human body follows a log-normal distribution centered at 4, the initial state is defined as

[0133] For a better understanding of the present invention, it should be noted that the respiration value in the present invention refers to the average number of respirations per minute; the heartbeat value refers to the average number of heartbeats per minute.

[0134] Embodiment 2

[0135] The present invention also discloses a medium on which a computer program is stored. When the computer program is executed by a processor, the functions of the millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation described in the first embodiment are realized.

[0136] Embodiment 3

[0137] The present invention also discloses a millimeter wave radar vital signs detection device based on SVMD and harmonic wave estimation, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the functions of the millimeter wave radar vital signs detection system based on SVMD and harmonic wave estimation in Embodiment 1 are realized.

[0138] The present invention uses a principal component analysis method based on mutual correlation to calculate the mutual correlation degree of phase information between the adjacent distance unit and the target distance unit, and on this basis, performs PCA on the phase signal of the unit to be measured, effectively utilizes the human vital sign signals dispersed in different distance units, and realizes the enhancement of breathing and heartbeat signals. When performing breathing and heartbeat estimation, the present invention uses the SVMD algorithm to decompose the enhanced vital sign signals, realizes the separation of breathing and heartbeat signals, selects IMF to reconstruct breathing and heartbeat signals based on the proportion of effective energy in total energy, performs spectrum peak estimation to obtain the corresponding frequencies of breathing and heartbeat and their harmonics, and realizes accurate estimation of breathing. The heartbeat and its harmonic frequencies are input into the Kalman filter, and the value closest to the predicted state is selected as the heartbeat estimation value, thereby realizing accurate estimation of the heartbeat.

[0139] The present invention is a measurement method for multi-distance unit energy integration and multi-heart rate harmonic joint estimation, which can extract vital sign signals dispersed in different distance units, enhance the global distribution characteristics of vital sign signals, and overcome the shortcomings of modal aliasing and parameter pre-definition in traditional modal decomposition methods. The present invention abandons the shortcomings of the VMD algorithm that the decomposition layer number K and the penalty factor α need to be pre-set, reduces the modal aliasing phenomenon existing in algorithms such as EMD, EEMD, and CEEMDAN, eliminates the interference of environmental clutter on respiratory and heartbeat signals, realizes the joint estimation of heartbeat and its harmonics, suppresses the influence of respiratory high-order harmonics on heartbeat estimation results, and improves the accuracy of respiratory and heartbeat estimation results.

[0140] In order to further illustrate the beneficial effects of the present invention, a simulation experiment is performed using the system in this embodiment. The final respiratory estimation value obtained by the simulation experiment is compared with the actual respiratory monitoring value. Figure 4 As shown in Figure 2, the final heart rate estimation value obtained through the simulation experiment and the actual heart rate monitoring value are compared. Figure 5 As shown. Figure 4 and Figure 5 It can be seen that the breathing and heartbeat estimation performed using the present invention is substantially consistent with the actual situation, thus proving the beneficial effects of the present invention.

[0141] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0142] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0143] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0144] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0145] Obviously, the above embodiments are merely examples for clear explanation and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived from these are still within the protection scope of the invention.

Claims

1. A millimeter wave radar vital signs detection system based on SVMD and harmonic wave estimation, characterized in that: include: The millimeter wave radar is placed in front of the human body to be detected and is used to collect the vital signs of the human body to obtain the IF signal; A target determination module is used to obtain the distance between the millimeter wave radar and the human body detected in the field of view according to the IF signal, and determine the position of the target human body in the area through a target detection algorithm; A candidate distance unit selection module is used to select a candidate distance unit adjacent to the target human body according to the phase information of the IF signal; A vital sign signal extraction module is used to perform principal component analysis on the phase information of the IF signal in the distance unit where the target human body is located and the phase information of the IF signal in the candidate distance unit, and extract the component with the largest variance as the vital sign signal; A signal reconstruction module, used for decomposing the vital sign signal and reconstructing a breathing signal, a heartbeat signal and a heartbeat harmonic signal; The estimation module is used to perform spectrum estimation and correction on the reconstructed breathing signal, heartbeat signal and heartbeat harmonic signal to obtain the final breathing estimation value and the final heartbeat estimation value.

2. The millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation according to claim 1 is characterized in that: The step of selecting a candidate distance unit adjacent to the target human body according to the phase information of the IF signal comprises: The overlapping sliding window algorithm is used to extract the phase information of the IF signal in the distance unit where the target human body is located and the phase information of the IF signal in the adjacent distance unit. The cross-correlation of the phase information of the IF signal in the distance unit where the target human body is located and the phase information of the IF signal in the adjacent distance unit is calculated, and the adjacent distance unit whose cross-correlation result is greater than the preset threshold is selected as the candidate distance unit.

3. The millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation according to claim 1 is characterized in that: Decomposing the vital sign signal includes: The vital sign signal is divided into a respiratory frequency component, a heartbeat frequency component and a noise frequency component, and the component that conforms to the respiratory frequency component is selected as the first mode, the classification that conforms to the heartbeat frequency component is selected as the second mode, and the component in high-quality harmonics is selected as the noise frequency component; The SVMD algorithm is used to extract the modes one by one until the reconstruction error reaches a certain threshold, and the target modal components of the respiratory frequency component and the heart rate frequency component are obtained.

4. The millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation according to claim 3 is characterized in that: When the SVMD algorithm is used to extract the modes one by one, The first part of the noise frequency component is 0, and the update method of the mode in the n+1th iteration is: in, is the estimated value of the Lth mode obtained in the n+1th iteration, L represents the number of decomposed modes, n represents the number of iterations, ω represents the current frequency, is the estimated value of the Lth mode obtained at the nth iteration, is the mapping of the original input signal in the frequency domain; is the center frequency of the Lth mode obtained in the nth iteration, ω i is the center frequency of the ith mode; is the Lagrange multiplier corresponding to ω, α represents the balance parameter of the data fidelity constraint, and α can be obtained by the Lagrange multiplier method.

5. The millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation according to claim 4 is characterized in that: The iterative update method of the center frequency of the Lth mode is: in, represents the center frequency of the Lth mode obtained in the n+1th iteration, is the estimated value of the Lth mode obtained at the n+1th iteration.

6. The millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation according to claim 5 is characterized in that: The iterative update method of Lagrange multipliers is: Among them, λ n+1 is the Lagrange multiplier obtained at the n+1th iteration, λ n is the Lagrange multiplier obtained in the nth iteration, τ represents the update parameter, is the estimated value of the ith mode obtained at the nth iteration, is the estimated value of the ith mode, is the value of the ith mode determined in the n+1th iteration, is the center frequency of the ith mode obtained in the n+1th iteration.

7. The millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation according to claim 3 is characterized in that: The reconstruction obtains a respiratory signal, a heartbeat signal and a heartbeat harmonic signal. Specifically, after selecting the optimal parameter group of the target modal components of the respiratory frequency component and the heartbeat frequency component, the vital sign signal is modally decomposed to obtain an intrinsic modal function, and the reconstructed respiratory signal, the reconstructed heartbeat fundamental frequency signal, the reconstructed heartbeat second harmonic signal and the reconstructed heartbeat third harmonic signal are obtained according to the intrinsic modal function.

8. The millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation according to claim 7 is characterized in that: The reconstruction process of the reconstructed respiratory signal is specifically as follows: Perform FFT analysis on each intrinsic mode function to obtain the total energy of each intrinsic mode function component; Calculate the total energy of each intrinsic mode function component and the total energy within the respiratory frequency band, and construct the judgment condition as follows: Where E(i) is the total energy of each intrinsic mode function component, E r (i) is the total energy within the respiratory frequency band, is an empirical parameter; if the judgment condition is met, the intrinsic modal function component is considered to be an effective component of the respiratory signal, and the intrinsic modal function components c1(t), c2(t),…, c of all respiratory signals that meet the judgment condition are selected. j (t),…,c j (t), c j (t) is the intrinsic modal function component of the jth respiratory signal that meets the judgment condition at time t, and k is the number of intrinsic modal function components of the respiratory signal that meets the judgment condition; The reconstructed respiratory signal is calculated as: d r =d r (1),d r (2),…,d r (t),…,d r (N), Among them, d r is the reconstructed respiratory signal, d r (t) is the reconstructed respiratory signal at time t, and N is the total time.

9. The millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation according to claim 7, characterized in that: The final breathing estimate is obtained by: The reconstructed breathing signal is subjected to fast Fourier transform to convert the time domain signal into a frequency domain signal, and the peak value of the frequency domain signal is selected as the final breathing estimation value.

10. The millimeter wave radar vital sign detection system based on SVMD and harmonic wave estimation according to claim 7, characterized in that: The final heartbeat estimation value is obtained by: Performing fast Fourier transform on the reconstructed heartbeat fundamental frequency signal, the reconstructed heartbeat second harmonic signal, and the reconstructed heartbeat third harmonic signal to convert the time domain signal into a frequency domain signal, and selecting the peak value of the frequency domain signal as the corresponding heartbeat fundamental frequency estimation frequency, heartbeat second harmonic estimation frequency, and heartbeat third harmonic estimation frequency; The estimated frequency of the heartbeat fundamental frequency, the estimated frequency of the heartbeat second harmonic, and the estimated frequency of the heartbeat third harmonic are used as the input of the Kalman filter, and the measurement vector is defined as: Among them, z h is the measurement vector, f h1 is the estimated frequency of the heartbeat fundamental frequency, f h2 is the estimated frequency of the second harmonic of the heartbeat, f h3 is the estimated frequency of the third harmonic of the heartbeat, and T represents the transposition operation; The state vector is constructed as: x=[f h f′ h f″ h ] T , Among them, x is the state vector, f h =f h1 , f′ h f h1 The first derivative of h f h1 The second derivative of According to the current state of the mth sliding window, the predicted state is calculated using the state transfer equation: in, is the current state of the mth sliding window, is the predicted state in the m+1th sliding window, F is the state transfer matrix, and Δt is the duration of the non-overlapping part of the sliding window; The state vector is represented as The covariance is P m,m The Gaussian process of , the predicted covariance is: P m+1,2 =FP m,m F T +Q, Among them, P m+1,m is the predicted covariance, Q is the covariance of the system process noise, is the process noise; Construct the observation matrix, denoted as H, Construct the measurement error, denoted as e m , Construct the relevant covariance, denoted as S m , S m =HP m+1,m H T +R; where R is a diagonal matrix containing the uncertainty of each heartbeat estimate region, is the expected square of the estimated error of the heartbeat fundamental frequency, is the expected square of the estimated error of the estimated frequency of the second harmonic of the heartbeat, is the expected square of the estimated error of the estimated frequency of the third harmonic of the heartbeat; Initialize the predefined Q and R, the initial state is The covariance is P 0,0 If there is a heartbeat estimate close to the predicted state Then in z h Select the state close to the prediction The heartbeat estimate is taken as the final heartbeat estimate; the distance between the new measurement vector and the filter prediction state is minimized, and the state vector is updated to z is the index corresponding to the selected heartbeat estimate in the measurement vector, e m (z) represents e m The zth element in k m is the Kalman gain, h z is the transpose of the zth row of H, S m (z,z) is S m The z-th diagonal element of ; the relevant covariance update for the next filter iteration, denoted as P m+1,m+1 , I is the identity matrix.

Images