Millimeter-wave radar breathing and heart rate synchronization monitoring method based on sparse spatiotemporal constraints

Abstract

The invention discloses a method for synchronous monitoring of breathing and heart rate by millimeter wave radar based on sparse spatiotemporal constraints, comprising: obtaining the echo pulse signal of the millimeter wave radar, obtaining the breathing signal and the heart rate signal, and obtaining the sampling echo signal matrix; decomposing the signal matrix into a low-rank matrix and a sparse matrix, introducing a total variation regularization term, and establishing a joint optimization objective function; iteratively optimizing the objective function, obtaining a low-rank matrix, a breathing signal sparse matrix, and a heart rate signal sparse matrix, and realizing synchronous monitoring of breathing and heart rate. Synchronous detection of apnea events and abnormal heart rate events is realized. The definition of total variation regularization is introduced in the establishment of the optimization model, which improves the decoupling and reconstruction accuracy of apnea and abnormal heart rate events. In the process of optimizing the solution algorithm, a feature selection process is added, the input data is pre-corrected, low-correlation features are screened out, and the robustness of the judgment of apnea and abnormal heart rate events is improved.

Description

Technical Field

[0001] The invention belongs to the technical field of respiratory monitoring, and relates to a method for synchronously monitoring respiratory and heart rate using a millimeter wave radar based on sparse spatiotemporal constraints. Background Art

[0002] With the development of non-contact monitoring technology, millimeter-wave radar has gradually become an effective tool for physiological signal monitoring due to its superior performance. Millimeter-wave radar can penetrate clothing and provide real-time, non-invasive monitoring, avoiding the discomfort and usage restrictions brought by traditional contact devices. In addition, millimeter-wave radar technology has strong anti-interference capabilities and can work stably in noisy environments. It is suitable for a variety of application scenarios such as hospitals, homes and public places.

[0003] However, existing millimeter-wave radar systems can usually only process single signals independently when monitoring respiration and heart rate, and lack effective algorithm models to simultaneously analyze these two physiological signals, resulting in the need to detect abnormal heart rate and apnea independently, making it impossible to detect potential health problems in a timely manner. Recent studies have shown that there is a close physiological connection between heart rate and respiration, and their changes often occur at the same time, so a new monitoring method is needed to combine the analysis and detection of these two signals. Summary of the invention

[0004] The purpose of the present invention is to provide a millimeter-wave radar breathing and heart rate synchronization monitoring method based on sparse spatiotemporal constraints, introduce the definition of total variation regularization in the establishment of the optimization model, and improve the decoupling and reconstruction accuracy of apnea and abnormal heart rate events.

[0005] The technical solution to achieve the purpose of the present invention is:

[0006] A method for synchronously monitoring breathing and heart rate using a millimeter wave radar based on sparse spatiotemporal constraints comprises the following steps:

[0007] S01: Acquire the echo pulse signal of the millimeter wave radar, obtain the breathing signal and the heart rate signal, and obtain the sampled echo signal matrix;

[0008] S02: Decompose the signal matrix into a low-rank matrix and a sparse matrix, introduce the total variation regularization term, and establish a joint optimization objective function;

[0009] S03: Iteratively optimize the objective function to obtain a low-rank matrix, a respiratory signal sparse matrix, and a heart rate signal sparse matrix to achieve synchronous monitoring of respiration and heart rate.

[0010] In the preferred technical solution, the method for obtaining the sampled echo signal matrix in step S01 includes:

[0011] Millimeter wave radar The transmitted signal of a pulse is:

[0012]

[0013] in, is the amplitude of the transmitted signal, is the initial frequency, is the frequency modulation bandwidth, is the modulation period;

[0014] After being reflected by the human body surface, the echo pulse signal received by the millimeter wave radar is expressed as:

[0015]

[0016] in, is the time delay of the transmitted signal reflected by the human body, and , is the radial distance between the radar system and the human body, is the speed of light;

[0017] The time domain form of the respiratory signal is:

[0018]

[0019] in, is the amplitude of the breathing signal, is the phase shift caused by breathing;

[0020] The time domain form of the heart rate signal is:

[0021]

[0022] in, is the amplitude of the heart rate signal, is the phase shift caused by the heartbeat;

[0023] The echo pulse signal is represented as the superposition of two components:

[0024]

[0025] in is the noise term;

[0026] Receiving Signals The sampling signal The sampling length is , the millimeter wave radar receives a total of The received pulse sampling signals form a sampling echo signal matrix:

[0027] .

[0028] In the preferred technical solution, the method for establishing the joint optimization objective function in step S02 includes:

[0029] Decompose the sampled echo signal matrix into:

[0030]

[0031] in, is a low-rank matrix, is a sparse matrix, is the noise matrix;

[0032] The joint optimization objective function is:

[0033]

[0034] in, is the kernel function of the matrix, It is a matrix norm, is the Frobenius norm of the matrix, is the sparse matrix corresponding to the apnea event, is the sparse matrix corresponding to abnormal heart rate events, With A regularization parameter that controls the strength of sparsity, represents the total variation regularization, and is the weight coefficient of the regularization term that controls the total variation;

[0035]

[0036]

[0037] in, and Represents sparse matrices and No. Line The elements of the column, The number of pulse sampling signals received, is the sampling length.

[0038] In the preferred technical solution, the method for iteratively optimizing the objective function includes:

[0039] S11: Extract time domain, frequency domain, and instantaneous features from the received signal matrix, and construct a corresponding feature matrix for each received pulse signal , whose dimensions are , is the number of features extracted;

[0040] S12: Use L1 regularization for feature selection to filter out features with high correlation with the target signal;

[0041] S13: Use the alternating multiplier algorithm to iteratively optimize and obtain the low-rank matrix estimate , sparse matrix estimate and .

[0042] In the preferred technical solution, the method of using L1 regularization for feature selection in step S12 includes:

[0043] S21: Define apnea target signature , which represents the time domain, frequency domain, and instantaneous characteristics of the received signal when a typical apnea event occurs; defines the target characteristic signal of abnormal heart rate event , which represents the time domain, frequency domain, and instantaneous characteristics of the received signal when an abnormal heart rate event occurs;

[0044] S22: For apnea events, define The apnea event feature weight vector corresponding to the received pulse , the model is optimized by solving the following features:

[0045]

[0046] Get based on The weight vector estimate of ;

[0047] S23: Define apnea event characteristic thresholds , the estimated value of the weight vector Perform threshold judgment. If it is greater than the threshold, it is retained, otherwise it is set to zero, that is:

[0048]

[0049] For all received After the feature weight vectors corresponding to the pulse signals are reconstructed and estimated, a new feature weight matrix is ​​formed. ;

[0050] S24: For abnormal heart rate events, define The characteristic weight vector of abnormal heart rate events corresponding to the received pulses , the model is optimized by solving the following features:

[0051]

[0052] Get based on The weight vector estimate of ;

[0053] S25: Define the characteristic threshold of abnormal heart rate events , the estimated value of the weight vector Perform threshold judgment. If it is greater than the threshold, it is retained, otherwise it is set to zero, that is:

[0054]

[0055] For all received After the feature weight vectors corresponding to the pulse signals are reconstructed and estimated, a new feature weight matrix is ​​formed. ;

[0056] S26: Update the reconstructed input signal after feature selection :

[0057] .

[0058] In the preferred technical solution, step S13 further includes reconstructing the input signal after selecting the updated feature Rewrite the joint optimization objective function as:

[0059] .

[0060] In the preferred technical solution, the sparse matrix estimate obtained based on the optimization solution is and , analyze and record apnea events and abnormal heart rate events through threshold judgment or modal decomposition.

[0061] The present invention also discloses a millimeter wave radar breathing and heart rate synchronization monitoring system based on sparse space-time constraints, comprising:

[0062] The acquisition module obtains the echo pulse signal of the millimeter wave radar, obtains the breathing signal and the heart rate signal, and obtains the sampled echo signal matrix;

[0063] The joint optimization objective function building module decomposes the signal matrix into a low-rank matrix and a sparse matrix, introduces a total variation regularization term, and establishes a joint optimization objective function;

[0064] The iterative optimization module iteratively optimizes the objective function to obtain a low-rank matrix, a respiratory signal sparse matrix, and a heart rate signal sparse matrix to achieve synchronous monitoring of respiration and heart rate.

[0065] In the preferred technical solution, the method for establishing the joint optimization objective function in the joint optimization objective function construction module includes:

[0066] Decompose the sampled echo signal matrix into:

[0067]

[0068] in, is a low-rank matrix, is a sparse matrix, is the noise matrix;

[0069] The joint optimization objective function is:

[0070]

[0071] in, is the kernel function of the matrix, It is a matrix norm, is the Frobenius norm of the matrix, is the sparse matrix corresponding to the apnea event, is the sparse matrix corresponding to abnormal heart rate events, and is the regularization parameter that controls the strength of sparsity, represents the total variation regularization, and is the weight coefficient of the regularization term that controls the total variation;

[0072]

[0073]

[0074] in, and Represents sparse matrices and No. Line The elements of the column, The number of pulse sampling signals received, is the sampling length.

[0075] The present invention further discloses a computer storage medium on which a computer program is stored. When the computer program is executed, the above-mentioned millimeter wave radar breathing and heart rate synchronization monitoring method based on sparse spatiotemporal constraints is implemented.

[0076] Compared with the prior art, the present invention has the following significant advantages:

[0077] The present invention realizes the synchronous detection of apnea events and abnormal heart rate events. The definition of total variation regularization is introduced in the establishment of the optimization model, which improves the decoupling and reconstruction accuracy of apnea and abnormal heart rate events. In the process of optimizing the solution algorithm, a feature selection process is added to pre-correct the input data, filter out low-correlation features, and improve the robustness of apnea and abnormal heart rate event determination. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 It is a flow chart of the method for synchronously monitoring breathing and heart rate of millimeter wave radar based on sparse spatiotemporal constraints of the present invention;

[0079] Figure 2 It is a principle block diagram of the millimeter wave radar breathing and heart rate synchronization monitoring system based on sparse space-time constraints of the present invention. DETAILED DESCRIPTION

[0080] The principle of the present invention is as follows: the present invention realizes the synchronous detection of apnea events and abnormal heart rate events. The definition of total variation regularization is introduced in the establishment of the optimization model, which improves the decoupling and reconstruction accuracy of apnea and abnormal heart rate events. In the process of optimizing the solution algorithm, a feature selection process is added, the input data is pre-corrected, low-correlation features are screened out, and the robustness of the judgment of apnea and abnormal heart rate events is improved.

[0081] Embodiment 1:

[0082] like Figure 1 As shown, a method for synchronously monitoring breathing and heart rate using a millimeter wave radar based on sparse spatiotemporal constraints comprises the following steps:

[0083] S01: Acquire the echo pulse signal of the millimeter wave radar, obtain the breathing signal and the heart rate signal, and obtain the sampled echo signal matrix;

[0084] S02: Decompose the signal matrix into a low-rank matrix and a sparse matrix, introduce the total variation regularization term, and establish a joint optimization objective function;

[0085] S03: Iteratively optimize the objective function to obtain a low-rank matrix, a respiratory signal sparse matrix, and a heart rate signal sparse matrix to achieve synchronous monitoring of respiration and heart rate.

[0086] In a preferred embodiment, the method of obtaining the sampled echo signal matrix in step S01 includes:

[0087] Millimeter wave radar The transmitted signal of a pulse is:

[0088]

[0089] in, is the amplitude of the transmitted signal, is the initial frequency, is the frequency modulation bandwidth, is the modulation period;

[0090] After being reflected by the human body surface, the echo pulse signal received by the millimeter wave radar is expressed as:

[0091]

[0092] in, is the time delay of the transmitted signal reflected by the human body, and , is the radial distance between the radar system and the human body, is the speed of light;

[0093] The time domain form of the respiratory signal is:

[0094]

[0095] in, is the amplitude of the breathing signal, is the phase shift caused by breathing;

[0096] The time domain form of the heart rate signal is:

[0097]

[0098] in, is the amplitude of the heart rate signal, is the phase shift caused by the heartbeat;

[0099] The echo pulse signal is represented as the superposition of two components:

[0100]

[0101] in is the noise term;

[0102] Receiving Signals The sampling signal The sampling length is , the millimeter wave radar receives a total of The received pulse sampling signals form a sampling echo signal matrix:

[0103] .

[0104] In a preferred embodiment, the method of establishing the joint optimization objective function in step S02 includes:

[0105] Decompose the sampled echo signal matrix into:

[0106]

[0107] in, is a low-rank matrix, is a sparse matrix, is the noise matrix;

[0108] The joint optimization objective function is:

[0109]

[0110] in, is the kernel function of the matrix, It is a matrix norm, is the Frobenius norm of the matrix, is the sparse matrix corresponding to the apnea event, is the sparse matrix corresponding to abnormal heart rate events, and is the regularization parameter that controls the strength of sparsity, represents the total variation regularization, and is the weight coefficient of the regularization term that controls the total variation;

[0111]

[0112]

[0113] in, and Represents sparse matrices and No. Line The elements of the column, The number of pulse sampling signals received, is the sampling length.

[0114] In a preferred embodiment, the method for iteratively optimizing the objective function includes:

[0115] S11: Extract time domain, frequency domain, and instantaneous features from the received signal matrix, and construct a corresponding feature matrix for each received pulse signal , whose dimensions are , is the number of features extracted;

[0116] S12: Use L1 regularization for feature selection to filter out features with high correlation with the target signal;

[0117] S13: Use the alternating multiplier algorithm to iteratively optimize and obtain the low-rank matrix estimate , sparse matrix estimate and .

[0118] In a preferred embodiment, the method of using L1 regularization to perform feature selection in step S12 includes:

[0119] S21: Define apnea target signature , which represents the time domain, frequency domain, and instantaneous characteristics of the received signal when a typical apnea event occurs; defines the target characteristic signal of abnormal heart rate event , which represents the time domain, frequency domain, and instantaneous characteristics of the received signal when an abnormal heart rate event occurs;

[0120] S22: For apnea events, define The apnea event feature weight vector corresponding to the received pulse , the model is optimized by solving the following features:

[0121]

[0122] Get based on The weight vector estimate of ;

[0123] S23: Define apnea event characteristic thresholds , the estimated value of the weight vector Perform threshold judgment. If it is greater than the threshold, it is retained, otherwise it is set to zero, that is:

[0124]

[0125] For all received After the feature weight vectors corresponding to the pulse signals are reconstructed and estimated, a new feature weight matrix is ​​formed. ;

[0126] S24: For abnormal heart rate events, define The characteristic weight vector of abnormal heart rate events corresponding to the received pulses , the model is optimized by solving the following features:

[0127]

[0128] Get based on The weight vector estimate of ;

[0129] S25: Define the characteristic threshold of abnormal heart rate events , the estimated value of the weight vector Perform threshold judgment. If it is greater than the threshold, it is retained, otherwise it is set to zero, that is:

[0130]

[0131] For all received After the feature weight vectors corresponding to the pulse signals are reconstructed and estimated, a new feature weight matrix is ​​formed. ;

[0132] S26: Update the reconstructed input signal after feature selection :

[0133] .

[0134] In a preferred embodiment, step S13 further comprises reconstructing the input signal after selecting the updated feature Rewrite the joint optimization objective function as:

[0135] .

[0136] In a preferred embodiment, the sparse matrix estimate obtained based on the optimization solution and , analyze and record apnea events and abnormal heart rate events through threshold judgment or modal decomposition.

[0137] In another embodiment, a computer storage medium stores a computer program, which, when executed, implements the above-mentioned millimeter-wave radar breathing and heart rate synchronization monitoring method based on sparse spatiotemporal constraints.

[0138] The millimeter-wave radar breathing and heart rate synchronization monitoring method based on sparse space-time constraints can adopt any of the above-mentioned millimeter-wave radar breathing and heart rate synchronization monitoring methods based on sparse space-time constraints, and the specific implementation will not be repeated here.

[0139] In another embodiment, if Figure 2 As shown, a millimeter wave radar breathing and heart rate synchronization monitoring system based on sparse space-time constraints includes:

[0140] The acquisition module 10 acquires the echo pulse signal of the millimeter wave radar, obtains the breathing signal and the heart rate signal, and obtains the sampled echo signal matrix;

[0141] A joint optimization objective function building module 20 decomposes the signal matrix into a low-rank matrix and a sparse matrix, introduces a total variation regularization term, and establishes a joint optimization objective function;

[0142] The iterative optimization module 30 iteratively optimizes the objective function to obtain a low-rank matrix, a respiratory signal sparse matrix and a heart rate signal sparse matrix, thereby realizing synchronous monitoring of respiration and heart rate.

[0143] Specifically, the workflow of the millimeter wave radar breathing and heart rate synchronization monitoring system based on sparse spatiotemporal constraints is described below by taking a preferred embodiment as an example:

[0144] In continuous pulse mode, millimeter wave radar detects the movement of the target by continuously transmitting and receiving a series of frequency modulated continuous wave (FMCW) signals, and identifies the breathing and heart rate characteristics of the human body by analyzing the phase or frequency changes of the return signal. FMCW radar transmits a linear frequency modulation signal during the detection process, and its frequency increases linearly with time. The frequency offset (i.e., Doppler shift) of the return signal is related to the distance and speed of the target and can be calculated through spectrum analysis. For respiratory monitoring, millimeter wave radar uses the phase change of the reflected signal to sense the tiny periodic movement of the human body surface, such as the expansion and contraction of the chest cavity. In heart rate monitoring, the beating of the heart also causes a tiny displacement of the chest, which has a higher frequency. Millimeter wave radar captures the above two types of signals at the same time and monitors breathing and heart rate by analyzing the frequency changes of the reflected signal.

[0145] Millimeter wave radar The transmitted signal of a pulse is:

[0146] (1)

[0147] in is the amplitude of the transmitted signal, is the initial frequency, is the frequency modulation bandwidth, is the modulation period.

[0148] After being reflected by the human body surface, the echo pulse signal received by the radar can be expressed as

[0149] (2)

[0150] in is the time delay of the transmitted signal reflected by the human body, and , is the radial distance between the radar system and the human body, The speed of light.

[0151] The respiratory rate is usually between 0.1 Hz and 0.5 Hz (i.e., 6 to 30 breaths per minute). The respiratory signal is mainly manifested as a periodic change at a relatively low frequency, and its waveform is usually smooth. The pulse signal reflects the periodic expansion and contraction of the chest, and its time domain form can be expressed as:

[0152] (3)

[0153] in, is the amplitude of the breathing signal, is the phase shift caused by breathing.

[0154] Heart rate frequencies are typically between 0.8 Hz and 2 Hz (i.e., 48 to 120 beats per minute). Heart rate signals have higher frequencies, change more rapidly, and usually have more significant amplitude changes in the respiratory signal. The time domain form of a received pulse signal can be expressed as:

[0155] (4)

[0156] in is the amplitude of the heart rate signal, The phase offset caused by the heartbeat.

[0157] When monitoring respiration and heart rate simultaneously, the radar system's echo pulse signal is represented by the sum of two components:

[0158] (5)

[0159] in is the noise term.

[0160] make Indicates receiving signal The sampling length of the signal is , then the millimeter wave radar system receives a total of The received pulse sampling signals can form a sampling echo signal matrix

[0161] (6)

[0162] The present invention represents the signal matrix collected by the millimeter wave radar as two parts:

[0163] 1. Low-rank part: represents normal, periodic physiological signals (including breathing and heart rate).

[0164] 2. Sparse part: Indicates abnormal physiological events (such as apnea, abnormal heart rate). These sudden events usually appear as sparse and irregular signals.

[0165] The above model can effectively separate the normal state from the abnormal events in the signal and detect apnea events and abnormal heart rate events more accurately. Specifically, the sampled echo signal matrix can be decomposed into the following model:

[0166] (7)

[0167] in is a low-rank matrix, representing normal physiological signals. It is a sparse matrix, representing sudden apnea events and abnormal heart rate events. is the noise matrix, representing the modeled measurement errors and environmental noise.

[0168] The present invention monitors the respiration and heart rate synchronously by using the sparse matrix of formula (7) With low rank matrix By separating the above two types of matrices, we can further monitor and analyze sudden events such as apnea and abnormal heart rate. The present invention proposes the following joint optimization problem to achieve the above separation. The specific joint optimization problem is:

[0169] (8)

[0170] in, is the sparse matrix corresponding to the apnea event, is the sparse matrix corresponding to abnormal heart rate events, and is the regularization parameter that controls the strength of sparsity, represents total variation regularization (TotalVariation), and is the weight coefficient of the regularization term that controls the total variation.

[0171] The specific definition of total variation is:

[0172] (9)

[0173] (10)

[0174] The starting point for introducing the total variation regularization term into the joint optimization problem in the present invention is that apnea events and abnormal heart rate events have a certain continuity in a short period of time, so the time consistency characteristics of apnea events can be better captured, thereby reducing misjudgment.

[0175] The solution process of the joint optimization problem shown in formula (8) can be summarized as the following steps:

[0176] Step 1: Extract time domain, frequency domain, and instantaneous features from the received signal matrix, and construct a corresponding feature matrix for each received pulse signal , whose dimensions are , is the number of features extracted.

[0177] Step 2: Use L1 regularization to perform feature selection to filter out features that are highly correlated with the target signal (respiration and heart rate).

[0178] Step 2.1: Define the apnea target signature , which represents the time domain, frequency domain, and instantaneous characteristics of the received signal when a typical apnea event occurs. Similarly, the target characteristic signal of abnormal heart rate event is defined , which represents the time domain, frequency domain, and instantaneous characteristics of the received signal when an abnormal heart rate event occurs.

[0179] Step 2.2: For apnea events, define The apnea event feature weight vector corresponding to the received pulse , by solving the following feature optimization model

[0180] (11)

[0181] Get based on The weight vector estimate of .

[0182] Step 2.3: Define apnea event feature thresholds , the estimated value of the weight vector obtained in step 2.2 Perform threshold judgment. If it is greater than the threshold, it is retained, otherwise it is set to zero.

[0183] (12)

[0184] Through the above method, all received After the feature weight vectors corresponding to the pulse signals are reconstructed and estimated, a new feature weight matrix is ​​formed. .

[0185] Step 2.4: For abnormal heart rate events, define The characteristic weight vector of abnormal heart rate events corresponding to the received pulses , by solving the following feature optimization model

[0186] (13)

[0187] Get based on The weight vector estimate of .

[0188] Step 2.5: Define the characteristic threshold of abnormal heart rate events , the estimated value of the weight vector obtained in step 2.4 Perform threshold judgment. If it is greater than the threshold, it is retained, otherwise it is set to zero.

[0189] (14)

[0190] Through the above method, all received After the feature weight vectors corresponding to the pulse signals are reconstructed and estimated, a new feature weight matrix is ​​formed. .

[0191] Step 2.6: Update the reconstructed input signal after feature selection

[0192] (15)

[0193] Step 3: Solve the following update optimization problem using the alternating multiplier (ADMM) algorithm:

[0194] (16)

[0195] And output the low-rank matrix estimate obtained by optimization solution , sparse matrix estimate and .

[0196] Finally, based on the sparse matrix obtained by optimization and , apnea events and abnormal heart rate events can be analyzed and recorded through existing technologies such as threshold judgment or modal decomposition.

[0197] The synchronous detection of apnea events and abnormal heart rate events is realized. The definition of total variation regularization is introduced in the establishment of the optimization model, which improves the accuracy of decoupling and reconstruction of apnea and abnormal heart rate events. In the process of optimizing the solution algorithm, a feature selection process is added to pre-correct the input data, filter out low-correlation features, and improve the robustness of apnea and abnormal heart rate event judgment.

[0198] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. A method for synchronous monitoring of breathing and heart rate using millimeter wave radar based on sparse spatiotemporal constraints, characterized in that: The following steps are involved: S01: Acquire the echo pulse signal of the millimeter wave radar, obtain the breathing signal and the heart rate signal, and obtain the sampled echo signal matrix; S02: Decompose the signal matrix into a low-rank matrix and a sparse matrix, introduce the total variation regularization term, and establish a joint optimization objective function; Methods for establishing joint optimization objective functions include: Decompose the sampled echo signal matrix into: ， in, is a low-rank matrix, is a sparse matrix, is the noise matrix; The joint optimization objective function is: ， in, is the kernel function of the matrix, It is a matrix norm, is the Frobenius norm of the matrix, is the sparse matrix corresponding to the apnea event, is the sparse matrix corresponding to abnormal heart rate events, and is the regularization parameter that controls the strength of sparsity, represents the total variation regularization, and is the weight coefficient of the regularization term that controls the total variation; ， ， in, and Represents sparse matrices and No. Line The elements of the column, is the number of pulse sampling signals received, is the sampling length; Methods for iterative optimization of the objective function include: S11: Extract time domain, frequency domain, and instantaneous features from the received signal matrix, and construct a corresponding feature matrix for each received pulse signal , whose dimensions are , is the number of features extracted; S12: Use L1 regularization for feature selection to screen out features that are highly correlated with the target signal; methods for using L1 regularization for feature selection include: S21: Define apnea target signature , which represents the time domain, frequency domain, and instantaneous characteristics of the received signal when a typical apnea event occurs; defines the target characteristic signal of abnormal heart rate event , which represents the time domain, frequency domain, and instantaneous characteristics of the received signal when an abnormal heart rate event occurs; S22: For apnea events, define The apnea event feature weight vector corresponding to the received pulse , the model is optimized by solving the following features: ， Get based on The weight vector estimate of ; S23: Define apnea event characteristic thresholds , the estimated value of the weight vector Perform threshold judgment. If it is greater than the threshold, it is retained, otherwise it is set to zero, that is: ， For all received After the feature weight vectors corresponding to the pulse signals are reconstructed and estimated, a new feature weight matrix is ​​formed. ; S24: For abnormal heart rate events, define The characteristic weight vector of abnormal heart rate events corresponding to the received pulses , the model is optimized by solving the following features: ， Get based on The weight vector estimate of ; S25: Define the characteristic threshold of abnormal heart rate events , the estimated value of the weight vector Perform threshold judgment. If it is greater than the threshold, it is retained, otherwise it is set to zero, that is: ， For all received After the feature weight vectors corresponding to the pulse signals are reconstructed and estimated, a new feature weight matrix is ​​formed. ; S26: Update the reconstructed input signal after feature selection : ； S13: Use the alternating multiplier algorithm to iteratively optimize and obtain the low-rank matrix estimate , sparse matrix estimate and ; S03: Iteratively optimize the objective function to obtain a low-rank matrix, a respiratory signal sparse matrix, and a heart rate signal sparse matrix to achieve synchronous monitoring of respiration and heart rate.

2. The method for synchronously monitoring breathing and heart rate using millimeter wave radar based on sparse spatiotemporal constraints according to claim 1, characterized in that: The method for obtaining the sampled echo signal matrix in step S01 includes: Millimeter wave radar The transmitted signal of a pulse is: ， in, is the amplitude of the transmitted signal, is the initial frequency, is the frequency modulation bandwidth, is the modulation period; After being reflected by the human body surface, the echo pulse signal received by the millimeter wave radar is expressed as: ， in, is the time delay of the transmitted signal reflected by the human body, and , is the radial distance between the radar system and the human body, is the speed of light; The time domain form of the respiratory signal is: ， in, is the amplitude of the breathing signal, is the phase shift caused by breathing; The time domain form of the heart rate signal is: ， in, is the amplitude of the heart rate signal, is the phase shift caused by the heartbeat; The echo pulse signal is represented as the superposition of two components: ， in is the noise term; Receiving Signals The sampling signal The sampling length is , the millimeter wave radar receives a total of The received pulse sampling signals form a sampling echo signal matrix: 。 3. The method for synchronously monitoring breathing and heart rate using millimeter wave radar based on sparse spatiotemporal constraints according to claim 1, characterized in that: Step S13 also includes reconstructing the input signal after selecting the updated feature Rewrite the joint optimization objective function as: 。 4. The method for synchronously monitoring breathing and heart rate using millimeter wave radar based on sparse spatiotemporal constraints according to claim 1, characterized in that: Sparse matrix estimates obtained based on optimization and , analyze and record apnea events and abnormal heart rate events through threshold judgment or modal decomposition.

5. A millimeter wave radar breathing and heart rate synchronization monitoring system based on sparse spatiotemporal constraints, characterized in that: include: The acquisition module obtains the echo pulse signal of the millimeter wave radar, obtains the breathing signal and the heart rate signal, and obtains the sampled echo signal matrix; The joint optimization objective function building module decomposes the signal matrix into a low-rank matrix and a sparse matrix, introduces a total variation regularization term, and establishes a joint optimization objective function; Methods for establishing joint optimization objective functions include: Decompose the sampled echo signal matrix into: ， in, is a low-rank matrix, is a sparse matrix, is the noise matrix; The joint optimization objective function is: ， in, is the kernel function of the matrix, It is a matrix norm, is the Frobenius norm of the matrix, is the sparse matrix corresponding to the apnea event, is the sparse matrix corresponding to abnormal heart rate events, and is the regularization parameter that controls the strength of sparsity, represents the total variation regularization, and is the weight coefficient of the regularization term that controls the total variation; ， ， in, and Represents sparse matrices and No. Line The elements of the column, is the number of pulse sampling signals received, is the sampling length; Methods for iterative optimization of the objective function include: S11: Extract time domain, frequency domain, and instantaneous features from the received signal matrix, and construct a corresponding feature matrix for each received pulse signal , whose dimensions are , is the number of features extracted; S12: Use L1 regularization for feature selection to screen out features that are highly correlated with the target signal; methods for using L1 regularization for feature selection include: S21: Define apnea target signature , which represents the time domain, frequency domain, and instantaneous characteristics of the received signal when a typical apnea event occurs; defines the target characteristic signal of abnormal heart rate event , which represents the time domain, frequency domain, and instantaneous characteristics of the received signal when an abnormal heart rate event occurs; S22: For apnea events, define The apnea event feature weight vector corresponding to the received pulse , the model is optimized by solving the following features: ， Get based on The weight vector estimate of ; S23: Define apnea event characteristic thresholds , the estimated value of the weight vector Perform threshold judgment. If it is greater than the threshold, it is retained, otherwise it is set to zero, that is: ， For all received After the feature weight vectors corresponding to the pulse signals are reconstructed and estimated, a new feature weight matrix is ​​formed. ; S24: For abnormal heart rate events, define The characteristic weight vector of abnormal heart rate events corresponding to the received pulses , the model is optimized by solving the following features: ， Get based on The weight vector estimate of ; S25: Define the characteristic threshold of abnormal heart rate events , the estimated value of the weight vector Perform threshold judgment. If it is greater than the threshold, it is retained, otherwise it is set to zero, that is: ， For all received After the feature weight vectors corresponding to the pulse signals are reconstructed and estimated, a new feature weight matrix is ​​formed. ; S26: Update the reconstructed input signal after feature selection : ； S13: Use the alternating multiplier algorithm to iteratively optimize and obtain the low-rank matrix estimate , sparse matrix estimate and ; The iterative optimization module iteratively optimizes the objective function to obtain a low-rank matrix, a respiratory signal sparse matrix, and a heart rate signal sparse matrix to achieve synchronous monitoring of respiration and heart rate.

6. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed, the method for synchronously monitoring respiration and heart rate by millimeter-wave radar based on sparse spatiotemporal constraints described in any one of claims 1 to 4 is implemented.

Images