Millimeter wave radar breath and heart rate synchronous monitoring method and system

Abstract

The invention relates to the field of heart rate monitoring, and discloses a millimeter wave radar breath and heart rate synchronous monitoring method and system, and the method comprises the steps: transmitting a linear frequency modulation continuous wave signal according to a millimeter wave radar, and collecting original echo data reflected by a target region; baseband signal demodulation and phase information extraction are carried out on the original echo data to obtain original phase time sequence data, and the original phase time sequence data comprise thoracic cavity micro-motion features; and according to a Butterworth band-pass filter, preprocessing the original phase time sequence data through human body physiological signal frequency band characteristics to obtain a breathing frequency band signal and a heart rate frequency band signal. According to the method, the apnea event triggering threshold value and the arrhythmia early warning index are updated in real time through Kalman filtering and extended Kalman filtering, so that the monitoring system can dynamically adjust the health parameters, which means that the monitoring system can be optimized in real time and the abnormal health event can be accurately responded in different physiological states.

Description

Technical Field

[0001] The present invention relates to the technical field of heart rate monitoring, and in particular to a method and system for synchronously monitoring respiration and heart rate using millimeter-wave radar. Background Art

[0002] Millimeter-wave radar is a radar technology that uses millimeter waves for detection and measurement. It analyzes the target's position, speed, shape, movement and other information by emitting millimeter-wave signals and receiving signals reflected from the target.

[0003] Existing methods have the following shortcomings: Traditional methods mostly rely on contact sensors (such as ECGs and respiratory belts) to acquire physiological signals. These methods are susceptible to interference from factors such as noise, motion artifacts, and poor contact, resulting in unstable signals and low accuracy. Furthermore, in traditional methods, signals of respiration, heart rate, and vascular pulsation often interfere with each other, making separation of these different physiological signals difficult. Traditional methods often rely on simple filtering techniques, but these techniques may not be able to effectively distinguish and decouple these complex signals. Furthermore, traditional methods often only analyze data based on fixed monitoring parameters and cannot dynamically adjust in real time to changes in physiological state. For example, the trigger thresholds for heart rate and apnea events may vary depending on individual differences or physiological state changes, but traditional methods struggle to achieve adaptive adjustment. Furthermore, traditional methods often rely on manual interpretation and diagnosis, requiring medical personnel to manually analyze and judge the data. This method is susceptible to human factors, resulting in diagnostic delays and low accuracy. Furthermore, traditional methods are susceptible to environmental noise. For example, the signals of ECG and respiratory sensors are susceptible to electromagnetic interference and other noise sources, resulting in unstable signal quality.

[0004] Therefore, the present application now proposes a millimeter wave radar breathing and heart rate synchronous monitoring method and system to solve the above problems. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for synchronously monitoring respiration and heart rate using millimeter-wave radar, comprising: According to the millimeter wave radar, a linear frequency modulated continuous wave signal is emitted to collect raw echo data reflected from the target area; baseband signal demodulation and phase information extraction are performed on the raw echo data to obtain raw phase time series data, wherein the raw phase time series data includes chest micro-motion characteristics; Preprocessing the original phase time series data using a Butterworth bandpass filter with the frequency band characteristics of human physiological signals to obtain a respiratory frequency band signal and a heart rate frequency band signal; determining pure heart rate phase micro-motion data based on the respiratory frequency band signal and the heart rate frequency band signal; and decoupling the pure heart rate phase micro-motion data into a fundamental frequency component related to ventricular ejection and a higher harmonic component related to vascular pulsation based on variational mode decomposition; Determining a heart rate variability time series model based on the phase continuity of the fundamental frequency component; performing spatiotemporal alignment on the respiratory frequency band signal and the heart rate variability time series model; performing physiological parameter fusion estimation on the respiratory frequency band signal and the heart rate variability time series model using a Kalman filter to obtain a fusion estimation result; and obtaining an apnea event triggering threshold and an arrhythmia warning indicator based on the fusion estimation result; The apnea event trigger threshold and the arrhythmia warning indicator are updated in real time according to the extended Kalman filter to obtain a dynamic health monitoring parameter set; the dynamic health monitoring parameter set is input into a pre-trained long short-term memory network, and respiratory pattern classification and arrhythmia type identification are performed according to the long short-term memory network to obtain respiratory pattern classification results and arrhythmia type identification results.

[0006] Preferably, the method further comprises: The beamforming weight of the millimeter-wave radar is adjusted according to the breathing pattern classification result, and the spatial spectrum of the abnormal breathing area is enhanced; the selective amplification of the characteristic frequency band is triggered according to the arrhythmia type identification result, and the micro-Doppler characteristic signal-to-noise ratio is optimized; the optimized dynamic health monitoring parameter set is encapsulated into a structured data stream, and the structured data stream is transmitted to the medical terminal device for visualization and abnormal alarm according to the serial communication protocol.

[0007] Preferably, determining the pure heart rate phase micro-motion data according to the respiratory frequency band signal and the heart rate frequency band signal includes: Performing Hilbert transform on the respiratory frequency band signal to obtain respiratory phase modulation envelope data; Performing short-time Fourier transform on the heart rate frequency band signal to obtain a time-frequency spectrum; determining characteristic points of cardiac periodicity according to the time-frequency spectrum diagram; determining a chest cavity displacement time function according to the respiratory phase modulation envelope data; According to the dynamic time warping algorithm, respiratory harmonic interference suppression is performed on the cardiac periodic feature points through the chest displacement time function to obtain pure heart rate phase micro-motion data.

[0008] Preferably, suppressing respiratory harmonic interference of the cardiac periodic feature points by using the chest displacement time function according to a dynamic time warping algorithm to obtain pure heart rate phase micro-motion data includes: A mapping relationship between respiratory motion and radar phase change and the chest displacement time function is established based on nonlinear fitting to generate a continuous chest vibration displacement curve; Constructing a multidimensional cost matrix based on the time axis deviation, the respiratory phase change and the heart rate frequency band energy fluctuation, wherein the multidimensional cost matrix includes a time dimension weight, a phase dimension weight and a spectrum dimension weight; The time dimension weight is adaptively adjusted based on the statistical characteristics of the respiratory cycle, giving priority to aligning the extreme points of the cardiac cycle and respiratory phase; the phase dimension weight is used to set a dynamic threshold according to the amplitude of chest displacement to suppress the phase folding effect caused by respiratory movement; the spectrum dimension weight is used to introduce energy distribution information of the time spectrum graph; Mapping the cardiac periodic characteristic points to a stable respiratory motion interval to generate corrected respiratory phase modulation envelope data after respiratory harmonics are suppressed; The corrected respiratory phase modulation envelope data is non-uniformly resampled to obtain time-axis aligned corrected respiratory phase modulation envelope data; the respiratory phase coupling degree corresponding to the corrected respiratory phase modulation envelope data is determined, and a multidimensional cost matrix is ​​adjusted according to the respiratory phase coupling degree. A balance between respiratory harmonic suppression and heart rate characteristics is achieved through iterative optimization to obtain pure heart rate phase micro-fluctuation data.

[0009] Preferably, decoupling the pure heart rate phase micro-fluctuation data into a fundamental frequency component related to ventricular ejection and a high-order harmonic component related to vascular pulsation according to variational mode decomposition includes: Performing micro-Doppler feature decomposition on the pure heart rate phase micro-motion data to obtain a time-frequency feature parameter set; wherein the time-frequency feature parameter set includes heart valve opening and closing features; The pure heart rate phase micro-fluctuation data is decoupled into a fundamental frequency component related to ventricular ejection and a high-order harmonic component related to vascular pulsation through the time-frequency characteristic parameter set according to variational mode decomposition.

[0010] Preferably, performing micro-Doppler eigendecomposition on the pure heart rate phase micro-motion data to obtain time-varying frequency components includes: Performing time-frequency analysis on the pure heart rate phase micro-motion data according to synchronous compression wavelet transform to match the characteristic frequency band of heart valve vibration; Performing cluster analysis on the characteristic frequency bands of the heart valve vibration according to the energy of the time-frequency spectrum to locate the instantaneous frequency mutation area, wherein the instantaneous frequency mutation area corresponds to the micro-Doppler impulse response generated by the opening and closing of the heart valve; Performing micro-Doppler eigendecomposition on the pure heart rate phase micro-motion data corresponding to the instantaneous frequency mutation region according to an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode functions, wherein the intrinsic mode functions include high-frequency transient components; performing envelope spectrum analysis on the multiple intrinsic mode functions to screen out characteristic components associated with heart valve motion to obtain a target intrinsic mode function; performing Hilbert-Huang transform on the target intrinsic mode function to obtain an instantaneous frequency curve; The frequency change rate of the instantaneous frequency curve is calculated according to the sliding window method; the time-frequency characteristic parameter set of the valve opening and closing event is determined according to the frequency change rate, wherein the time-frequency characteristic parameter set includes the opening and closing interval time, the frequency modulation depth and the instantaneous frequency slope.

[0011] Preferably, performing physiological parameter fusion estimation through the respiratory frequency band signal and the heart rate variability time series model according to the Kalman filter to obtain a fusion estimation result includes: The respiratory frequency band signal after time-space alignment is used as the state input; the heart rate variability time series model after time-space alignment is used as the observation input; the chest displacement amplitude caused by breathing is set as the first state variable, and the instantaneous phase of the heart rate variability is set as the first observation variable; Determine a first state equation based on the first state variable, the first process noise and the state transfer matrix of the respiratory rhythm, and determine a first observation equation based on the first observation variable, the first observation noise and the respiratory heart rate coupling matrix; An initial state covariance is set according to the amplitude variance of the respiratory frequency band signal; a first process noise covariance is determined according to the stability of the respiratory frequency band signal; and a first observation noise covariance is determined according to the phase jitter characteristics of the heart rate variability time series model; Predicting the current respiratory state based on the state estimate at the previous moment and the first state equation; updating the state prediction covariance; fusing the confidence of the respiratory prediction and the heart rate variability observation; correcting the respiratory prediction based on the first observation variable; updating the state covariance; and outputting the optimal fused state at the current moment; A fusion estimation result is obtained according to the optimal fusion state at the current moment, wherein the fusion estimation result includes a respiratory waveform and a heart rate variability time series that decouples respiratory interference.

[0012] Preferably, the apnea event trigger threshold and the arrhythmia warning indicator are updated in real time according to the extended Kalman filter to obtain a dynamic health monitoring parameter set, including: The fusion estimation result is used as input data; the apnea event trigger threshold is set as a second state variable; and the arrhythmia warning indicator is set as a third state variable; Determine a second state equation based on the second state variable, the third state variable, the nonlinear threshold evolution function, and the second process noise; determine a second observation equation based on the respiratory frequency band signal, the heart rate variability time series model, the nonlinear observation function, and the second observation noise; Determining a respiratory initial value based on the historical amplitude distribution of the respiratory waveform; determining a heart rate variability initial value based on the local standard deviation of the heart rate variability time series; determining a second process noise covariance based on the fluctuation range of the individual physiological parameter; determining a second observation noise covariance based on the signal-to-noise ratio of the respiratory waveform and the heart rate variability time series decoupled from respiratory interference; The current threshold is predicted according to the state at the previous moment; the Jacobian matrix is ​​calculated; the predicted covariance is updated according to the Jacobian matrix; and the minimum value of the sliding window amplitude of the respiratory waveform and the difference between adjacent RR intervals of the heart rate variability time series are used as the second observation vector.

[0013] Preferably, the apnea event trigger threshold and the arrhythmia warning indicator are updated in real time according to the extended Kalman filter to obtain a dynamic health monitoring parameter set, further comprising: Determine a Jacobian measurement matrix based on the current threshold; determine a Kalman gain based on the Jacobian measurement matrix; and fuse the observation values ​​to correct the apnea event trigger threshold and the arrhythmia warning indicator; Detecting apnea according to the corrected apnea event trigger threshold to obtain an apnea event sequence; evaluating heart rhythm according to the corrected arrhythmia warning indicator to obtain a heart rhythm state sequence; A dynamic health monitoring parameter set is acquired according to the apnea event sequence and the heart rhythm state sequence.

[0014] A millimeter-wave radar respiration and heart rate synchronous monitoring system, which is applicable to the above-mentioned millimeter-wave radar respiration and heart rate synchronous monitoring method, comprises: a phase extraction unit configured to collect raw echo data reflected from a target area based on a linear frequency modulated continuous wave signal emitted by the millimeter-wave radar; perform baseband signal demodulation and phase information extraction on the raw echo data to obtain raw phase time series data, wherein the raw phase time series data includes chest micro-motion characteristics; A modal decomposition unit is configured to preprocess the original phase time series data using a Butterworth bandpass filter through frequency band characteristics of human physiological signals to obtain respiratory frequency band signals and heart rate frequency band signals; determine pure heart rate phase micro-motion data based on the respiratory frequency band signals and the heart rate frequency band signals; and decouple the pure heart rate phase micro-motion data into a fundamental frequency component related to ventricular ejection and a higher harmonic component related to vascular pulsation based on variational modal decomposition; a fusion estimation unit, the fusion estimation unit being configured to determine a heart rate variability time series model based on the phase continuity of a fundamental frequency component; perform spatiotemporal alignment on the respiratory frequency band signal and the heart rate variability time series model; perform fusion estimation of physiological parameters using the respiratory frequency band signal and the heart rate variability time series model according to a Kalman filter to obtain a fusion estimation result; and obtain an apnea event triggering threshold and an arrhythmia warning indicator based on the fusion estimation result; a health monitoring unit configured to update the apnea event trigger threshold and the arrhythmia warning indicator in real time according to an extended Kalman filter to obtain a dynamic health monitoring parameter set; input the dynamic health monitoring parameter set into a pre-trained long short-term memory network, and perform respiratory pattern classification and arrhythmia type identification according to the long short-term memory network to obtain respiratory pattern classification results and arrhythmia type identification results; An abnormal alarm unit is used to adjust the beamforming weight of the millimeter-wave radar according to the breathing pattern classification result, and perform spatial spectrum enhancement on the abnormal breathing area; trigger the selective amplification of the characteristic frequency band according to the arrhythmia type identification result, and optimize the micro-Doppler characteristic signal-to-noise ratio; encapsulate the optimized dynamic health monitoring parameter set into a structured data stream, and transmit the structured data stream to the medical terminal device according to the serial communication protocol for visualization and abnormal alarm.

[0015] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention can monitor the chest micro-motion characteristics of the target area in real time by using the reflected signal obtained by the millimeter wave radar. These micro-motion information can reflect physiological parameters related to breathing and heart rate, providing strong technical support for accurately assessing human health status. Moreover, through the application of variational mode decomposition, the pure heart rate phase micro-motion data can be effectively decoupled into fundamental frequency components and high-order harmonic components, thereby accurately distinguishing different physiological signals and avoiding signal cross-interference, which makes the monitoring of heart rate and vascular pulsation more accurate. (2) The present invention uses Kalman filtering and extended Kalman filtering to update the apnea event trigger threshold and arrhythmia warning indicator in real time, so that the monitoring system can dynamically adjust health parameters. This means that the monitoring system can be optimized in real time under different physiological conditions and accurately respond to abnormal health events. In addition, by using Kalman filtering to perform physiological parameter fusion estimation, the respiratory frequency band signal and the heart rate variability time series model can be combined to achieve data fusion of multiple signal sources. This fusion can obtain more reliable health monitoring results in a noisy environment. (3) By using a long short-term memory network to classify and identify breathing patterns and arrhythmia types, the present invention can learn and predict the user's health trends based on historical data, further improving the intelligence and automation level of diagnosis. Based on the breathing pattern classification and arrhythmia identification results, the system can dynamically adjust the beamforming weight of the millimeter-wave radar and perform spatial spectrum enhancement on the abnormal breathing area, thereby improving the detection accuracy of the target area. In addition, the selective amplification of the characteristic frequency band and the optimization of the micro-Doppler characteristic signal-to-noise ratio help to improve the quality and accuracy of the radar signal in complex environments. (4) The present invention encapsulates the optimized dynamic health monitoring parameter set into a structured data stream and transmits it to the medical terminal device through a serial communication protocol for visualization and abnormal alarm. This enables medical personnel to obtain accurate health monitoring data in real time and promptly discover and respond to abnormal health conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of the steps of the overall method in one embodiment of the present invention; Figure 2 FIG. 1 is a schematic diagram of the system architecture structure of the overall system in one embodiment of the present invention.

[0017] In the figure: 1. Phase extraction unit; 2. Modal decomposition unit; 3. Fusion estimation unit; 4. Health monitoring unit; 5. Abnormal alarm unit. DETAILED DESCRIPTION

[0018] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0019] For example 1, please refer to Figure 1 The present invention provides a technical solution: a method for synchronously monitoring breathing and heart rate using a millimeter-wave radar, comprising: S1. The millimeter-wave radar transmits a linear frequency-modulated continuous wave signal and collects raw echo data reflected from the target area; baseband signal demodulation and phase information extraction are performed on the raw echo data to obtain raw phase time series data, wherein the raw phase time series data includes chest micro-motion characteristics; S2. Preprocessing the original phase time series data using the frequency band characteristics of human physiological signals using a Butterworth bandpass filter to obtain respiratory frequency band signals and heart rate frequency band signals; determining pure heart rate phase micro-fluctuation data based on the respiratory frequency band signals and heart rate frequency band signals; and decoupling the pure heart rate phase micro-fluctuation data into a fundamental frequency component related to ventricular ejection and a higher-order harmonic component related to vascular pulsation based on variational mode decomposition; S3. Determine a heart rate variability time series model based on the phase continuity of the fundamental frequency component; perform spatiotemporal alignment of the respiratory frequency band signal and the heart rate variability time series model; perform physiological parameter fusion estimation using the respiratory frequency band signal and the heart rate variability time series model using a Kalman filter to obtain a fusion estimation result; obtain an apnea event trigger threshold and an arrhythmia warning indicator based on the fusion estimation result; S4. Update the apnea event trigger threshold and arrhythmia warning indicator in real time according to the extended Kalman filter to obtain a dynamic health monitoring parameter set; input the dynamic health monitoring parameter set into a pre-trained long short-term memory network, and perform respiratory pattern classification and arrhythmia type recognition according to the long short-term memory network to obtain respiratory pattern classification results and arrhythmia type recognition results.

[0020] It should be noted that the millimeter-wave radar transmits a linear frequency modulated continuous wave signal: the millimeter-wave radar transmits a continuous wave signal whose frequency changes linearly with time, and the signal is reflected back by the target object (such as the human body); the radar receives the reflected signal and forms echo data, which contains dynamic information of the human body; by demodulating the echo data, the phase information of the signal can be extracted; the phase information can reveal the tiny movements of the target, such as the micro-movement of the chest cavity; the micro-movement of the chest cavity reflects the human body's movements during breathing and heartbeat, so they are important features for monitoring breathing and heart rate; the Butterworth bandpass filter filters the phase time series data according to the frequency characteristics of human physiological signals (such as breathing frequency, heart rate frequency, etc.) to remove irrelevant frequency components; the purpose is to extract frequency band signals related to breathing and heartbeat; breathing frequency band signal: usually between 0.1 Hz and 0.5 Hz; heart rate frequency band signal: usually between 0.8 Hz and 2.5 Hz; pure heart rate phase micro-fluctuation data can be used to extract signals related to heart rate based on the filtered data; these signals can reflect each beat of the heart; variational mode decomposition (VMD) is used to decompose heart rate phase micro-fluctuation data into two important components: the fundamental frequency component is directly related to ventricular ejection (the pumping action of the heart); the higher harmonic component is related to vascular pulsation (such as the fluctuation of blood in the blood vessels); by analyzing the phase of the fundamental frequency component, the change of heart rate can be inferred; this is the basis of the heart rate variability (HRV) model; the heart rate variability time series model describes the change of heart rate over time and is used to monitor the regularity and abnormality of heart rate; the respiratory frequency band signal and the heart rate variability time series model are synchronized in time and space so that breathing and heart rate data can be considered simultaneously in the same time window; Kalman filtering is used to fuse the respiratory signal and the heart rate variability signal to estimate physiological parameters; this is a recursive filtering process The program can dynamically update the estimated value based on the existing measurement data and prediction results; the fused results can provide more accurate respiratory and heart rate parameters; the apnea event trigger threshold refers to when the system detects apnea, a threshold is set by fusing the estimation results to trigger an apnea warning; the arrhythmia warning indicator refers to the heart rate data estimated through fusion. If the heart rate is irregular (such as arrhythmia), the system will issue a warning; because events such as apnea and arrhythmia will change over time, the extended Kalman filter can update the trigger thresholds and warning indicators of these events in real time, making the health monitoring parameter set more dynamic and accurate; the real-time updated dynamic health monitoring parameter set is input into the pre-trained LSTM model for respiratory pattern classification and arrhythmia type identification; different types of respiratory patterns are classified (such as normal breathing, shallow breathing, etc.); and the types of arrhythmias such as atrial fibrillation and tachycardia are identified.

[0021] In an optional embodiment, the method further includes: S5. Adjust the beamforming weight of the millimeter-wave radar based on the breathing pattern classification results, and perform spatial spectrum enhancement on the abnormal breathing area; trigger the selective amplification of the characteristic frequency band based on the arrhythmia type identification results, and optimize the micro-Doppler characteristic signal-to-noise ratio; encapsulate the optimized dynamic health monitoring parameter set into a structured data stream, and transmit the structured data stream to the medical terminal device according to the serial communication protocol for visualization and abnormal alarm.

[0022] It should be noted that according to the breathing pattern classification results, the beamforming weight of the millimeter-wave radar can be adjusted to enhance signal acquisition in abnormal breathing areas, thereby improving monitoring accuracy; according to the arrhythmia type identification results, the characteristic frequency band related to arrhythmia is selectively amplified to optimize the signal-to-noise ratio of the micro-Doppler feature and improve signal quality; the optimized monitoring data is transmitted to the medical terminal device in the form of a structured data stream through a serial communication protocol for visual presentation and abnormal alarm, providing doctors or users with real-time health status feedback.

[0023] In an optional embodiment, determining the pure heart rate phase micro-motion data according to the respiratory frequency band signal and the heart rate frequency band signal includes: A1. Performing Hilbert transform on the respiratory frequency band signal to obtain respiratory phase modulation envelope data; A2. Perform short-time Fourier transform on the heart rate frequency band signal to obtain a time-frequency spectrum; A3. Determine the characteristic points of cardiac periodicity based on the time-frequency spectrum; A4. determining a chest cavity displacement time function based on respiratory phase modulation envelope data; A5. According to the dynamic time warping algorithm, respiratory harmonic interference of cardiac periodic feature points is suppressed through the chest displacement time function to obtain pure heart rate phase micro-motion data.

[0024] It should be noted that the Hilbert Transform is used to extract the envelope signal and phase information from the original signal; here, the envelope data of the respiratory frequency band signal obtained by the Hilbert Transform can reflect the amplitude modulation of the respiratory signal (that is, the impact of respiratory changes on other physiological signals); the Hilbert Transform is a mathematical operation that generates a signal with a 90-degree phase difference from the signal and combines it with the original signal to calculate the instantaneous amplitude and phase of the signal; the final "envelope signal" reflects the periodic changes of breathing; the Short-Time Fourier Transform (STFT) is used to analyze the spectral changes of the signal over time, and is particularly suitable for processing non-stationary signals; the heart rate signal is usually non-stationary, so the Short-Time Fourier Transform The Fourier transform can provide the frequency information of the signal changing with time; the short-time Fourier transform divides the signal into short time segments and performs Fourier transform on each segment to obtain a time-frequency spectrum; the time-frequency spectrum shows the frequency components of the heart rate signal at different time points, which can help analyze the cardiac cycle and its changes; the periodic characteristic points of the heartbeat are identified from the time-frequency spectrum, which usually refers to the peak value of the heart rate, representing a complete heartbeat cycle; these characteristic points are the signs of the cardiac cycle and are very important for further analysis; by observing the features in the time-frequency spectrum, the periodic change points of the signal are determined, usually these characteristic points correspond to the starting point, end point or specific local maximum of the heartbeat; accurate identification Identifying these points is the basis for extracting heart rate micro-motion data; chest displacement refers to the ups and downs of the chest with breathing, which is usually directly related to the amplitude change of the respiratory signal; the envelope data of the respiratory frequency band can be used to establish the time function of chest displacement, which represents the displacement of the chest at different time points; the phase modulation envelope data of the respiratory frequency band signal is used to infer the time change of chest displacement; through a simple mathematical model or data fitting, a function that describes the change of chest displacement over time is obtained; the dynamic time warping algorithm (DTW) is a nonlinear sequence alignment technology, which is mainly used to deal with the alignment problem of time series signals; it can find the best matching path between time series of different lengths. path, eliminating temporal inconsistencies; through the chest displacement time function, the DTW algorithm can align the time misalignment between the respiratory signal and the heart rate signal, thereby effectively removing the interference of the respiratory frequency band signal on the heart rate signal (i.e., respiratory harmonic interference); through this process, pure heart rate phase micro-fluctuation data can be obtained; through the above processing and interference suppression, the final heart rate signal is no longer affected by the harmonics of breathing, retaining more accurate cardiac information; through the dynamic time warping algorithm, the interference component of the respiratory signal is removed from the heart rate signal, and the result is pure heart rate phase micro-fluctuation data; this data reflects the subtle changes in heart rate and can usually be used for further physiological analysis.

[0025] In an optional embodiment, respiratory harmonic interference suppression is performed on cardiac periodic feature points using a chest displacement time function according to a dynamic time warping algorithm to obtain pure heart rate phase micro-motion data, including: B1. Establish the mapping relationship between respiratory motion and radar phase change and the chest displacement time function based on nonlinear fitting to generate a continuous chest vibration displacement curve; B2. Constructing a multidimensional cost matrix based on the time axis deviation, respiratory phase change, and heart rate frequency band energy fluctuation, wherein the multidimensional cost matrix includes time dimension weight, phase dimension weight, and spectrum dimension weight; Among them, the time dimension weight is adaptively adjusted according to the statistical characteristics of the respiratory cycle, giving priority to aligning the extreme points of the cardiac cycle and respiratory phase; the phase dimension weight is used to set a dynamic threshold according to the amplitude of chest displacement to suppress the phase folding effect caused by respiratory movement; the spectrum dimension weight is used to introduce the energy distribution information of the time spectrum graph; B3, mapping the cardiac periodic characteristic points to the stable interval of respiratory movement to generate corrected respiratory phase modulation envelope data after respiratory harmonic suppression; B4. Perform non-uniform resampling on the corrected respiratory phase modulation envelope data to obtain time-axis aligned corrected respiratory phase modulation envelope data; determine the respiratory phase coupling corresponding to the corrected respiratory phase modulation envelope data, adjust the multidimensional cost matrix according to the respiratory phase coupling, and achieve a balance between respiratory harmonic suppression and heart rate characteristics through iterative optimization to obtain a pure heart rate phase micro-fluctuation number.

[0026] It should be noted that, through the nonlinear fitting method, the mapping relationship between respiratory movement and radar phase change is first established; this means that the phase change measured by the radar can be used to infer the displacement of the chest cavity; the chest displacement time function represents the vibration change of the chest cavity over a period of time, thereby providing data for subsequent analysis; in order to make the respiratory phase, heart rate and their relationship more accurate, a multidimensional cost matrix is ​​constructed, which contains three dimensions: time dimension, phase dimension and spectrum dimension; the time dimension weight analyzes the statistical characteristics of the respiratory cycle and dynamically adjusts the weight to ensure the alignment of the cardiac cycle and respiratory phase, especially at the extreme points of the cardiac cycle and respiratory phase; the phase dimension weight suppresses the phase folding effect caused by respiratory movement by dynamically setting the threshold; this is because the periodic change of breathing will affect the phase calculation in some cases, resulting in errors, so it needs to be adjusted through the phase dimension weight; the spectrum dimension weight introduces the time spectrum diagram (such as wavelet transform) for frequency analysis, and optimizes the breathing and heart rate through the spectrum dimension weight. relationship, eliminating noise; when performing respiratory phase modulation, harmonic interference caused by respiratory movement may occur, affecting the accurate extraction of heart rate; respiratory harmonics are suppressed by mapping cardiac periodic feature points to the stable interval of respiratory movement; this means that sampling and analysis should be performed in the area where breathing is stable, thereby reducing the impact of respiratory fluctuations on heart rate extraction; the suppressed respiratory phase modulation envelope data is non-uniformly resampled to align the data time axis, making subsequent analysis and optimization more accurate; respiratory phase coupling is calculated based on the corrected respiratory phase modulation envelope data. The higher the coupling, the stronger the mutual influence between respiration and heart rate; after calculating the respiratory phase coupling, the balance between respiratory and heart rate phases is gradually optimized by adjusting the weights of each dimension in the multidimensional cost matrix; this process is completed through iterative optimization, the purpose of which is to minimize the interference between heart rate characteristics and respiratory phase, and ultimately obtain pure heart rate phase micro-fluctuations, that is, eliminating the interference of respiratory movement on the heart rate signal.

[0027] In an optional embodiment, the pure heart rate phase micro-fluctuation data is decoupled into a fundamental frequency component related to ventricular ejection and a high-order harmonic component related to vascular pulsation according to variational mode decomposition, including: C1. Performing micro-Doppler feature decomposition on the pure heart rate phase micro-motion data to obtain a time-frequency feature parameter set; wherein the time-frequency feature parameter set includes heart valve opening and closing features; C2. Based on variational mode decomposition, the pure heart rate phase micro-fluctuation data is decoupled into the fundamental frequency component related to ventricular ejection and the high-order harmonic components related to vascular pulsation through the time-frequency characteristic parameter set.

[0028] It should be noted that heart rate phase micro-motion data refers to the tiny phase changes caused by heart contraction and relaxation, which are obtained by monitoring the heart's motion signals; usually, micro-motion data can be captured by high-precision sensors (such as radar, ultrasound, etc.), and these micro-motion data can reflect the subtle movements of the heart; pure heart rate phase refers to heart motion data that has been free of external interference and noise. This type of data focuses on changes related to heart activity and can provide detailed information on heart valve movement, heart wall movement, etc.; time-frequency features refer to the distribution characteristics of signals in the time and frequency domains, which are often obtained through time-frequency analysis methods (such as wavelet transform, short-time Fourier transform, etc.); these features are particularly important for analyzing non-stationary signals because heart signals have obvious time and frequency changes; the time-frequency feature parameter set includes various features that describe physiological phenomena such as the opening and closing of heart valves. These features may include parameters such as frequency, amplitude, and phase, which are used to characterize heart activity. Different physical phenomena in motion; in this application, VMD is used to decouple heart rate phase micro-motion data, separating the signal into different frequency components related to different physiological processes; the fundamental frequency component usually represents the main activity frequency of the heart and may be related to the process of ventricular ejection; this frequency is closely related to the contraction and relaxation of the ventricles with each heartbeat; the higher-order harmonic components usually correspond to the details of the heart movement, such as high-frequency components related to vascular pulsation and blood flow velocity; these components may be related to factors such as the elasticity of the vascular wall and the propagation of pulse waves; ventricular ejection is the process by which the heart pumps blood from the ventricles to the whole body, and its main frequency is usually the fundamental frequency of the heartbeat (heart rate); this frequency usually appears in the cardiac cycle and is directly related to the process of ventricular contraction; vascular pulsation is the vibration of the vascular wall driven by the blood flow caused by the heartbeat, and usually exhibits frequency components higher than the fundamental frequency; the higher-order harmonic components can reflect changes in blood flow, the elasticity of the blood vessels, and their response to pressure waves.

[0029] In an optional embodiment, micro-Doppler eigendecomposition is performed on the pure heart rate phase micro-motion data to obtain time-varying frequency components, including: D1. Perform time-frequency analysis on pure heart rate phase micro-motion data based on synchronous compression wavelet transform to match the characteristic frequency band of heart valve vibration; D2. Cluster analysis of characteristic frequency bands of heart valve vibration is performed based on the energy of the time-frequency spectrum to locate regions of instantaneous frequency mutations, which correspond to micro-Doppler impulse responses generated by the opening and closing of the heart valves. D3. performing micro-Doppler eigendecomposition on the pure heart rate phase micro-motion data corresponding to the instantaneous frequency mutation region according to an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode functions, wherein the intrinsic mode functions include high-frequency transient components; D4. Performing envelope spectrum analysis on multiple intrinsic mode functions to screen out characteristic components associated with heart valve motion to obtain a target intrinsic mode function; performing a Hilbert-Huang transform on the target intrinsic mode function to obtain an instantaneous frequency curve; D5. Calculate the frequency change rate of the instantaneous frequency curve according to the sliding window method; determine the time-frequency characteristic parameter set of the valve opening and closing event according to the frequency change rate, wherein the time-frequency characteristic parameter set includes the opening and closing interval time, the frequency modulation depth and the instantaneous frequency slope.

[0030] It should be noted that synchronized compressed wavelet transform (SWT) is used to perform time-frequency analysis on pure heart rate phase micro-motion data; wavelet transform is used to perform time-frequency analysis on the frequency band of heart valve vibration, thereby extracting the vibration characteristics of the heart valve; wavelet transform is used to decompose heart data into different frequency components, especially focusing on the vibration frequency range of the heart valve; through this method, the instantaneous frequency changes of valve vibration can be observed; by analyzing the energy distribution in the time-spectrum diagram, the unique vibration frequency band of the heart valve is identified and clustered; based on the energy information in the time-spectrum diagram (which shows the frequency distribution at different time points), cluster analysis is performed to locate the instantaneous frequency region of mutation; the mutation region may be related to the opening and closing action of the valve; the micro-Doppler impulse response characteristics in the instantaneous frequency mutation region are further analyzed; the pure heart rate phase micro-motion data corresponding to the mutation region are subjected to empirical mode decomposition to extract multiple intrinsic mode functions (IMFs); these IMFs represent different frequency components in the data, including high-frequency transient components, which may be related to the movement of the valve; envelope spectrum analysis is used to extract Characteristic components related to heart valve vibration; envelope spectrum analysis is performed on multiple intrinsic mode functions to screen out characteristic frequency bands related to heart valve movement (especially valve opening and closing); these frequency bands help to understand the physical process of valve vibration; further extract the instantaneous frequency curve of the target intrinsic mode function; perform Hilbert-Huang transform on the selected target intrinsic mode function to obtain the instantaneous frequency curve; this curve can describe the motion state of the valve and its frequency changes in detail; analyze the motion characteristics of the valve according to the rate of instantaneous frequency change; use the sliding window method to calculate the frequency change rate to obtain the dynamic characteristics of the valve motion; these change rates reflect the process of valve opening and closing; extract the time-frequency characteristics related to the valve opening and closing events through the frequency change rate; determine three key parameters through the frequency change rate: opening and closing interval time: the time required for the heart valve to transition from one state to another; frequency modulation depth: represents the amplitude of frequency change; instantaneous frequency slope: represents the rate of frequency change, revealing the rapid change characteristics of valve action.

[0031] In an optional embodiment, a physiological parameter fusion estimation is performed using a respiratory frequency band signal and a heart rate variability time series model according to a Kalman filter to obtain a fusion estimation result, including: E1. Use the spatiotemporally aligned respiratory frequency band signal as the state input; use the spatiotemporally aligned heart rate variability time series model as the observation input; set the chest displacement amplitude caused by breathing as the first state variable, and set the instantaneous phase of heart rate variability as the first observation variable; E2. Determine a first state equation based on the first state variable, the first process noise, and the state transfer matrix of the respiratory rhythm, and determine a first observation equation based on the first observation variable, the first observation noise, and the respiratory-heart rate coupling matrix; E3. Setting an initial state covariance based on the amplitude variance of the respiratory frequency band signal; determining a first process noise covariance based on the stability of the respiratory frequency band signal; and determining a first observation noise covariance based on the phase jitter characteristics of the heart rate variability time series model. E4. Predict the current respiratory state based on the state estimate at the previous moment and the first state equation; update the state prediction covariance; fuse the confidence of the respiratory prediction and the heart rate variability observation; correct the respiratory prediction based on the first observation variable; update the state covariance; and output the optimal fused state at the current moment; E5. Obtain a fusion estimation result based on the optimal fusion state at the current moment, wherein the fusion estimation result includes a respiratory waveform and a heart rate variability time series that decouples respiratory interference.

[0032] It should be noted that the state input refers to the respiratory frequency band signal after time and space alignment, which represents the changes in breathing; usually, in signal processing, this can be a signal related to breathing extracted by a filter (such as the amplitude of the respiratory frequency band); the observation input is a time series model of heart rate variability, which usually reflects the state of the autonomic nervous system through changes in heart rate (HRV); the time series model can include instantaneous phase changes of heart rate; the first state variable (amplitude of chest displacement caused by breathing) is used to describe the changes in chest displacement during breathing, reflecting the amplitude or intensity of breathing; the first observation variable is used to describe the instantaneous phase of heart rate variability, The instantaneous phase of heart rate variability reflects the changing characteristics of heart rate and can be used to indirectly analyze the regulatory effect of the autonomic nervous system on breathing; the state equation describes how the state variable (amplitude of chest displacement caused by breathing) changes over time, and is corrected by the state transfer matrix and process noise; process noise reflects unpredictable random disturbances in the model (such as natural fluctuations in breathing); the observation equation describes how the observed variable (instantaneous phase of heart rate variability) depends on the state variable, and is affected by observation noise and the coupling matrix; the coupling matrix reflects the relationship between breathing and heart rate variability; the initial state covariance is set according to the amplitude variance of the respiratory frequency band signal, which represents the uncertainty of the model about the initial state; the process noise covariance is set by the stability of the respiratory signal; if the respiratory signal is relatively stable, the process noise covariance is small, which means that the model's prediction of respiratory state changes is more accurate; the observation noise Acoustic covariance sets the observation noise covariance based on the phase jitter characteristics of the heart rate variability time series model, representing the degree of model uncertainty in the heart rate variability observation. A Kalman filter is used to predict the current state based on the state estimate and state equation at the previous moment. The prediction is based on the system's dynamic model and takes process noise into account. The covariance matrix is ​​updated to reflect the uncertainty of the current state estimate. The covariance matrix is ​​typically adjusted based on the introduction of new observations. By fusing information from different sources (respiration predictions and heart rate variability observations), the contribution of each observation to the state estimate is adjusted. Observations with higher confidence have a greater impact on the final result. The prediction result is corrected using the observed variable (the instantaneous phase of heart rate variability) to ensure more accurate state estimation. The ultimate goal is to output the optimal fusion state, which combines all information (respiration and heart rate variability signals) to achieve the best estimate. The final estimated results include: the respiratory waveform, which is the true respiratory signal estimated by the model; and the heart rate variability time series with decoupled respiratory interference, which is the heart rate signal obtained by removing the interference of respiration on heart rate variability, which can better reflect the heart's autonomous rhythm.

[0033] In an optional embodiment, the apnea event trigger threshold and the arrhythmia warning indicator are updated in real time according to the extended Kalman filter to obtain a dynamic health monitoring parameter set, including: F1. Use the fusion estimation result as input data; set the apnea event trigger threshold as the second state variable; set the arrhythmia warning indicator as the third state variable; F2, determining a second state equation based on the second state variable, the third state variable, the nonlinear threshold evolution function, and the second process noise; determining a second observation equation based on the respiratory frequency band signal, the heart rate variability time series model, the nonlinear observation function, and the second observation noise; F3. Determine the initial value of respiration based on the historical amplitude distribution of the respiratory waveform; determine the initial value of heart rate variability based on the local standard deviation of the heart rate variability time series; determine the second process noise covariance based on the fluctuation range of the individual physiological parameters; determine the second observation noise covariance based on the signal-to-noise ratio of the respiratory waveform and the heart rate variability time series decoupled from respiratory interference; F4. Predict the current threshold based on the state at the previous moment; calculate the Jacobian matrix; update the predicted covariance based on the Jacobian matrix; and use the minimum amplitude of the sliding window of the respiratory waveform and the difference between adjacent RR intervals of the heart rate variability time series as the second observation vector.

[0034] It should be noted that the data obtained from multiple data sources or sensors have undergone some form of fusion or processing and are used as input data for subsequent analysis; these data may include physiological signals such as respiratory waveforms, heart rate, blood oxygen, and movement; the second state variable refers to the apnea event trigger threshold, which usually means that when apnea exceeds a certain time, a warning or alarm is triggered; the third state variable refers to the arrhythmia warning indicator, which may involve the calculation of electrocardiogram or heart rate variability to detect abnormal heart rate; the second state equation is determined by the second state variable, the third state variable, the nonlinear threshold evolution function and the second process noise; this equation is used to describe the system state that changes with time, such as the dynamic changes of the apnea trigger threshold and the arrhythmia warning indicator; process noise refers to the randomness or non-randomness in the model Determinism is usually caused by fluctuations in the external environment or the system itself; based on the respiratory frequency band signal, heart rate variability time series model, nonlinear observation function and second observation noise, the second observation equation is defined; this equation characterizes the way to obtain the system state through observation data (such as respiratory signal, heart rate variability, etc.); observation noise refers to the error or uncertainty of the observation value, which may be caused by factors such as signal interference and sensor error; the respiratory value is initialized according to the historical amplitude distribution of the respiratory waveform; the heart rate variability value is initialized by the local standard deviation of the heart rate variability time series; secondly, the covariance of the process noise is set according to the fluctuation range of the individual physiological parameters, and the observation noise covariance is determined by the signal-to-noise ratio of the respiratory waveform and the heart rate variability time series that decouples the respiratory interference; the state and threshold at the current moment are predicted.

[0035] In an optional embodiment, the apnea event trigger threshold and the arrhythmia warning indicator are updated in real time according to the extended Kalman filter to obtain a dynamic health monitoring parameter set, further comprising: F5. Determine the Jacobian measurement matrix based on the current threshold; determine the Kalman gain based on the Jacobian measurement matrix; fuse the observation values ​​to correct the apnea event trigger threshold and arrhythmia warning indicator; F6. Detecting apnea based on the calibrated apnea event trigger threshold to obtain an apnea event sequence; evaluating the heart rhythm based on the calibrated arrhythmia warning indicator to obtain a heart rhythm state sequence; F7. Obtain a dynamic health monitoring parameter set based on the apnea event sequence and the heart rhythm state sequence.

[0036] It should be noted that the Jacobian matrix is ​​used to linearize the change of state and use it to update the predicted covariance; the Jacobian matrix is ​​used to calculate the local linear approximation of the nonlinear system; the minimum amplitude of the sliding window of the respiratory waveform and the difference between adjacent RR intervals of the heart rate variability time series are used as the second observation vector; these vectors provide important observation information to help judge the current health status; in the implementation process, the extended Kalman filter is used to update the apnea event trigger threshold and the arrhythmia warning indicator in real time; through the Kalman filter, the system state can be dynamically adjusted while constantly acquiring new observation data, so as to To obtain more accurate health monitoring parameters; through the Jacobian measurement matrix and Kalman gain, the observations are fused and the apnea event trigger threshold and arrhythmia warning indicator are corrected; the Kalman gain determines the weight of the new observation during correction, and a larger gain indicates that the new observation has a higher degree of confidence; the corrected apnea event trigger threshold is used to detect apnea events and generate an apnea event sequence; similarly, the corrected arrhythmia warning indicator is used to evaluate the heart rhythm and generate a heart rhythm state sequence; finally, based on the generated apnea event sequence and heart rhythm state sequence, a dynamic health monitoring parameter set is obtained. These parameter sets can be used to monitor and evaluate an individual's health status in real time.

[0037] For example 2, please refer to Figure 2 The present invention provides a technical solution: a millimeter wave radar respiration and heart rate synchronous monitoring system, which is applicable to the above-mentioned millimeter wave radar respiration and heart rate synchronous monitoring method, comprising: Phase extraction unit 1, which is used to collect original echo data reflected from the target area based on the linear frequency modulation continuous wave signal emitted by the millimeter wave radar; perform baseband signal demodulation and phase information extraction on the original echo data to obtain original phase time series data, wherein the original phase time series data includes chest micro-motion characteristics; Mode decomposition unit 2 is used to pre-process the original phase time series data according to the frequency band characteristics of the human physiological signal using a Butterworth bandpass filter to obtain a respiratory frequency band signal and a heart rate frequency band signal; determine pure heart rate phase micro-motion data based on the respiratory frequency band signal and the heart rate frequency band signal; and decouple the pure heart rate phase micro-motion data into a fundamental frequency component related to ventricular ejection and a higher harmonic component related to vascular pulsation according to variational mode decomposition; Fusion estimation unit 3, which is used to determine the heart rate variability time series model based on the phase continuity of the fundamental frequency component; align the respiratory frequency band signal and the heart rate variability time series model in time and space; perform physiological parameter fusion estimation through the respiratory frequency band signal and the heart rate variability time series model according to the Kalman filter to obtain a fusion estimation result; and obtain an apnea event triggering threshold and an arrhythmia warning indicator based on the fusion estimation result; Health monitoring unit 4, which is used to update the apnea event trigger threshold and arrhythmia warning indicator in real time based on the extended Kalman filter to obtain a dynamic health monitoring parameter set; input the dynamic health monitoring parameter set into a pre-trained long short-term memory network, and perform respiratory pattern classification and arrhythmia type recognition based on the long short-term memory network to obtain respiratory pattern classification results and arrhythmia type recognition results; Abnormal alarm unit 5, abnormal alarm unit 5 is used to adjust the beamforming weight of the millimeter wave radar according to the breathing pattern classification result, and perform spatial spectrum enhancement on the abnormal breathing area; trigger the selective amplification of the characteristic frequency band according to the arrhythmia type identification result, and optimize the micro-Doppler characteristic signal-to-noise ratio; encapsulate the optimized dynamic health monitoring parameter set into a structured data stream, and transmit the structured data stream to the medical terminal device according to the serial communication protocol for visualization and abnormal alarm.

[0038] The embodiments of the present invention are described in detail above with reference to the accompanying drawings, but the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A millimeter wave radar breathing and heart rate synchronous monitoring method, characterized in that ,include: The millimeter-wave radar transmits a linear frequency-modulated continuous wave signal and collects the original echo data reflected from the target area; Performing baseband signal demodulation and phase information extraction on the original echo data to obtain original phase time series data, wherein the original phase time series data includes chest micro-motion characteristics; Preprocessing the original phase time series data using a Butterworth bandpass filter with the frequency band characteristics of human physiological signals to obtain a respiratory frequency band signal and a heart rate frequency band signal; determining pure heart rate phase micro-motion data based on the respiratory frequency band signal and the heart rate frequency band signal; and decoupling the pure heart rate phase micro-motion data into a fundamental frequency component related to ventricular ejection and a higher harmonic component related to vascular pulsation based on variational mode decomposition; Determining a heart rate variability time series model based on the phase continuity of the fundamental frequency component; performing spatiotemporal alignment on the respiratory frequency band signal and the heart rate variability time series model; performing physiological parameter fusion estimation on the respiratory frequency band signal and the heart rate variability time series model using a Kalman filter to obtain a fusion estimation result; and obtaining an apnea event triggering threshold and an arrhythmia warning indicator based on the fusion estimation result; The apnea event trigger threshold and the arrhythmia warning indicator are updated in real time according to the extended Kalman filter to obtain a dynamic health monitoring parameter set; the dynamic health monitoring parameter set is input into a pre-trained long short-term memory network, and respiratory pattern classification and arrhythmia type identification are performed according to the long short-term memory network to obtain respiratory pattern classification results and arrhythmia type identification results.

2. A millimeter wave radar breathing and heart rate synchronous monitoring method according to claim 1, characterized in that The method further comprises: The beamforming weight of the millimeter-wave radar is adjusted according to the breathing pattern classification result, and the spatial spectrum of the abnormal breathing area is enhanced; the selective amplification of the characteristic frequency band is triggered according to the arrhythmia type identification result, and the micro-Doppler characteristic signal-to-noise ratio is optimized; the optimized dynamic health monitoring parameter set is encapsulated into a structured data stream, and the structured data stream is transmitted to the medical terminal device for visualization and abnormal alarm according to the serial communication protocol.

3. The method for synchronously monitoring respiration and heart rate using millimeter wave radar according to claim 2, wherein: Determining pure heart rate phase micro-motion data according to the respiratory frequency band signal and the heart rate frequency band signal includes: Performing Hilbert transform on the respiratory frequency band signal to obtain respiratory phase modulation envelope data; Performing short-time Fourier transform on the heart rate frequency band signal to obtain a time-frequency spectrum; determining characteristic points of cardiac periodicity according to the time-frequency spectrum diagram; determining a chest cavity displacement time function according to the respiratory phase modulation envelope data; According to the dynamic time warping algorithm, respiratory harmonic interference suppression is performed on the cardiac periodic feature points through the chest displacement time function to obtain pure heart rate phase micro-motion data.

4. The method for synchronously monitoring respiration and heart rate using millimeter wave radar according to claim 3, wherein: According to the dynamic time warping algorithm, respiratory harmonic interference suppression is performed on the cardiac periodic feature points through the chest displacement time function to obtain pure heart rate phase micro-motion data, including: A mapping relationship between respiratory motion and radar phase change and the chest displacement time function is established based on nonlinear fitting to generate a continuous chest vibration displacement curve; Constructing a multidimensional cost matrix based on the time axis deviation, respiratory phase change, and heart rate frequency band energy fluctuation, wherein the multidimensional cost matrix includes a time dimension weight, a phase dimension weight, and a spectrum dimension weight; The time dimension weight is adaptively adjusted based on the statistical characteristics of the respiratory cycle, giving priority to aligning the extreme points of the cardiac cycle and respiratory phase; the phase dimension weight is used to set a dynamic threshold according to the amplitude of chest displacement to suppress the phase folding effect caused by respiratory movement; the spectrum dimension weight is used to introduce energy distribution information of the time spectrum graph; Mapping the cardiac periodic characteristic points to a stable respiratory motion interval to generate corrected respiratory phase modulation envelope data after respiratory harmonics are suppressed; The corrected respiratory phase modulation envelope data is non-uniformly resampled to obtain time-axis aligned corrected respiratory phase modulation envelope data; the respiratory phase coupling degree corresponding to the corrected respiratory phase modulation envelope data is determined, and a multidimensional cost matrix is ​​adjusted according to the respiratory phase coupling degree. A balance between respiratory harmonic suppression and heart rate characteristics is achieved through iterative optimization to obtain pure heart rate phase micro-fluctuation data.

5. The method for synchronously monitoring respiration and heart rate using millimeter wave radar according to claim 4, wherein: Decoupling the pure heart rate phase micro-fluctuation data into a fundamental frequency component related to ventricular ejection and a high-order harmonic component related to vascular pulsation according to variational mode decomposition, including: Performing micro-Doppler feature decomposition on the pure heart rate phase micro-motion data to obtain a time-frequency feature parameter set; wherein the time-frequency feature parameter set includes heart valve opening and closing features; The pure heart rate phase micro-fluctuation data is decoupled into a fundamental frequency component related to ventricular ejection and a high-order harmonic component related to vascular pulsation through the time-frequency characteristic parameter set according to variational mode decomposition.

6. The method for synchronously monitoring respiration and heart rate using millimeter-wave radar according to claim 5, wherein: Performing micro-Doppler eigendecomposition on the pure heart rate phase micro-motion data to obtain time-varying frequency components, including: Performing time-frequency analysis on the pure heart rate phase micro-motion data according to synchronous compression wavelet transform to match the characteristic frequency band of heart valve vibration; Performing cluster analysis on the characteristic frequency bands of the heart valve vibration according to the energy of the time-frequency spectrum to locate the instantaneous frequency mutation area, wherein the instantaneous frequency mutation area corresponds to the micro-Doppler impulse response generated by the opening and closing of the heart valve; Performing micro-Doppler eigendecomposition on the pure heart rate phase micro-motion data corresponding to the instantaneous frequency mutation region according to an empirical mode decomposition algorithm to obtain a plurality of intrinsic mode functions, wherein the intrinsic mode functions include high-frequency transient components; performing envelope spectrum analysis on the multiple intrinsic mode functions to screen out characteristic components associated with heart valve motion to obtain a target intrinsic mode function; performing Hilbert-Huang transform on the target intrinsic mode function to obtain an instantaneous frequency curve; The frequency change rate of the instantaneous frequency curve is calculated according to the sliding window method; the time-frequency characteristic parameter set of the valve opening and closing event is determined according to the frequency change rate, wherein the time-frequency characteristic parameter set includes the opening and closing interval time, the frequency modulation depth and the instantaneous frequency slope.

7. The method for synchronously monitoring respiration and heart rate using millimeter wave radar according to claim 6, wherein: Performing physiological parameter fusion estimation using the respiratory frequency band signal and the heart rate variability time series model according to a Kalman filter to obtain a fusion estimation result, including: The respiratory frequency band signal after time-space alignment is used as the state input; the heart rate variability time series model after time-space alignment is used as the observation input; the chest displacement amplitude caused by breathing is set as the first state variable, and the instantaneous phase of the heart rate variability is set as the first observation variable; Determine a first state equation based on the first state variable, the first process noise and the state transfer matrix of the respiratory rhythm, and determine a first observation equation based on the first observation variable, the first observation noise and the respiratory heart rate coupling matrix; An initial state covariance is set according to the amplitude variance of the respiratory frequency band signal; a first process noise covariance is determined according to the stability of the respiratory frequency band signal; and a first observation noise covariance is determined according to the phase jitter characteristics of the heart rate variability time series model; Predicting the current respiratory state based on the state estimate at the previous moment and the first state equation; updating the state prediction covariance; fusing the confidence of the respiratory prediction and the heart rate variability observation; correcting the respiratory prediction based on the first observation variable; updating the state covariance; and outputting the optimal fused state at the current moment; A fusion estimation result is obtained according to the optimal fusion state at the current moment, wherein the fusion estimation result includes a respiratory waveform and a heart rate variability time series that decouples respiratory interference.

8. The method for synchronously monitoring respiration and heart rate using millimeter wave radar according to claim 7, wherein: The apnea event trigger threshold and the arrhythmia warning indicator are updated in real time according to the extended Kalman filter to obtain a dynamic health monitoring parameter set, including: The fusion estimation result is used as input data; the apnea event trigger threshold is set as a second state variable; and the arrhythmia warning indicator is set as a third state variable; Determine a second state equation based on the second state variable, the third state variable, the nonlinear threshold evolution function, and the second process noise; determine a second observation equation based on the respiratory frequency band signal, the heart rate variability time series model, the nonlinear observation function, and the second observation noise; Determining a respiratory initial value based on the historical amplitude distribution of the respiratory waveform; determining a heart rate variability initial value based on the local standard deviation of the heart rate variability time series; determining a second process noise covariance based on the fluctuation range of the individual physiological parameter; determining a second observation noise covariance based on the signal-to-noise ratio of the respiratory waveform and the heart rate variability time series decoupled from respiratory interference; The current threshold is predicted according to the state at the previous moment; the Jacobian matrix is ​​calculated; the predicted covariance is updated according to the Jacobian matrix; and the minimum value of the sliding window amplitude of the respiratory waveform and the difference between adjacent RR intervals of the heart rate variability time series are used as the second observation vector.

9. The method for synchronously monitoring respiration and heart rate using millimeter wave radar according to claim 8, wherein: The apnea event trigger threshold and the arrhythmia warning indicator are updated in real time according to the extended Kalman filter to obtain a dynamic health monitoring parameter set, further comprising: Determine a Jacobian measurement matrix based on the current threshold; determine a Kalman gain based on the Jacobian measurement matrix; and fuse the observation values ​​to correct the apnea event trigger threshold and the arrhythmia warning indicator; Detecting apnea according to the corrected apnea event trigger threshold to obtain an apnea event sequence; evaluating heart rhythm according to the corrected arrhythmia warning indicator to obtain a heart rhythm state sequence; A dynamic health monitoring parameter set is acquired according to the apnea event sequence and the heart rhythm state sequence.

10. A millimeter wave radar respiration and heart rate synchronization monitoring system, which is applicable to the millimeter wave radar respiration and heart rate synchronization monitoring method according to any one of claims 1 to 9, characterized in that: include: A phase extraction unit (1), the phase extraction unit (1) is used to collect original echo data reflected from the target area according to the linear frequency modulation continuous wave signal emitted by the millimeter wave radar; Performing baseband signal demodulation and phase information extraction on the original echo data to obtain original phase time series data, wherein the original phase time series data includes chest micro-motion characteristics; A modal decomposition unit (2) is used to pre-process the original phase time series data according to the frequency band characteristics of human physiological signals using a Butterworth bandpass filter to obtain a respiratory frequency band signal and a heart rate frequency band signal; determine pure heart rate phase micro-motion data according to the respiratory frequency band signal and the heart rate frequency band signal; and decouple the pure heart rate phase micro-motion data into a fundamental frequency component related to ventricular ejection and a high-order harmonic component related to vascular pulsation according to variational modal decomposition; A fusion estimation unit (3) is used to determine a heart rate variability time series model based on the phase continuity of the fundamental frequency component; perform spatiotemporal alignment on the respiratory frequency band signal and the heart rate variability time series model; perform physiological parameter fusion estimation on the respiratory frequency band signal and the heart rate variability time series model according to a Kalman filter to obtain a fusion estimation result; and obtain an apnea event triggering threshold and an arrhythmia warning indicator based on the fusion estimation result; A health monitoring unit (4), the health monitoring unit (4) is used to update the apnea event trigger threshold and the arrhythmia warning indicator in real time according to the extended Kalman filter to obtain a dynamic health monitoring parameter set; input the dynamic health monitoring parameter set into a pre-trained long short-term memory network, and perform respiratory pattern classification and arrhythmia type recognition according to the long short-term memory network to obtain a respiratory pattern classification result and an arrhythmia type recognition result; An abnormal alarm unit (5) is used to adjust the beamforming weight of the millimeter wave radar according to the breathing pattern classification result, and perform spatial spectrum enhancement on the abnormal breathing area; trigger the selective amplification of the characteristic frequency band according to the arrhythmia type identification result, and optimize the micro-Doppler characteristic signal-to-noise ratio; encapsulate the optimized dynamic health monitoring parameter set into a structured data stream, and transmit the structured data stream to the medical terminal device according to the serial communication protocol for visual presentation and abnormal alarm.

Images