Heart rate monitoring method, system and equipment based on millimeter wave radar and storage medium

Abstract

The invention discloses a heart rate monitoring method, system and device based on millimeter wave radar and a storage medium. The method comprises the following steps: extracting a phase signal of a target monitoring object in a radar echo signal; performing second-order differential processing to obtain an acceleration signal segment set corresponding to the phase signal; positioning an acceleration signal segment most similar to the periodic heartbeat signal from the acceleration signal segment set; segmenting beat-by-beat heartbeat signals from the positioned acceleration signal segments, and calculating the heart rate according to the segmented beat-by-beat heartbeat signals. According to the method, the acceleration signal segment of the distance angle bin with the optimal heartbeat is positioned based on the periodicity of the heartbeat signals, and the beat-by-beat heartbeat signals are segmented based on the acceleration signal segment, so that the accurate heart rate value is calculated, the misjudgment problem of a spectrum peak value detection method under the coexistence of multiple persons and the interference of respiration harmonics and body movement is effectively solved, and the accuracy of the detection result is improved. And the heart rate calculation precision is improved.

Description

Technical Field

[0001] The present invention relates to the field of biomedical signal processing technology, and in particular to a heart rate monitoring method, system, device and storage medium based on millimeter wave radar. Background Art

[0002] Vital sign monitoring radar transmits electromagnetic wave signals to the human body through a transmitter. When the signals reach the body's surface, tiny displacements of the chest wall caused by breathing and heartbeats modulate the phase of the reflected echo, causing the phase to carry information about the periodic motion of breathing and heartbeats. After denoising, time-frequency analysis is used to analyze the spectral characteristics of the echo phase signal, from which key physiological parameters such as respiratory rate and heart rate are separated.

[0003] Current heart rate measurement algorithms cannot achieve high-precision heart rate measurement due to weak heartbeat signals, dynamic interference coupling between respiratory and heartbeat signals, and signal crosstalk in multiple target scenarios that causes effective echo attenuation.

[0004] Therefore, the methods in the prior art need to be further improved. Summary of the Invention

[0005] In view of the deficiencies in the above-mentioned related technologies, the purpose of the present invention is to provide a heart rate monitoring method, system, device and storage medium based on millimeter wave radar to overcome the defect of the existing technology in lacking a method for highly accurate heart rate measurement.

[0006] The technical solutions adopted by the present invention to solve the technical problems are as follows:

[0007] In a first aspect, the present application provides a heart rate monitoring method based on millimeter wave radar, which includes:

[0008] Receive radar echo signals and extract the phase signal of the target monitoring object in the radar echo signals;

[0009] Processing the phase signal using second-order differential to obtain a set of acceleration signal segments corresponding to the phase signal;

[0010] Locating, from the set of acceleration signal segments, an acceleration signal segment that is most similar to a periodic heartbeat signal based on the heartbeat periodicity feature;

[0011] A beat-by-beat heartbeat signal is segmented from the located acceleration signal segment, and the heart rate is calculated based on the segmented beat-by-beat heartbeat signal.

[0012] Optionally, the step of receiving a radar echo signal and extracting a phase signal of a target monitoring object in the radar echo signal includes:

[0013] Perform fast Fourier transform on each chirp cycle in the radar echo signal in a fast time to obtain the distance-time matrix. Based on the distance-time matrix, the distance information of the target monitoring object is obtained.

[0014] The time average intensity of each pixel in the distance-time matrix is ​​used as the background noise to remove the static background noise signal from the distance-time matrix;

[0015] Calculate the covariance matrix of the target monitoring object's echo signal based on the distance information of the target monitoring object obtained by positioning, perform beamforming based on the covariance matrix, and separate multiple target signals;

[0016] The target monitoring object is located according to the different positions corresponding to the multiple separated target signals;

[0017] Perform arc tangent demodulation on the baseband signal corresponding to the located target monitoring object to obtain the phase signal of the target monitoring object.

[0018] Optionally, the step of processing the phase signal using second-order differential to obtain a set of acceleration signal segments corresponding to the phase signal includes:

[0019] The acceleration value corresponding to each sampling point in the phase signal is calculated respectively by using a differentiator based on least squares smoothing to obtain an acceleration signal corresponding to each sampling point in the phase signal. Each acceleration signal constitutes an acceleration signal segment set corresponding to the phase signal.

[0020] Optionally, the step of locating an acceleration signal segment that is most similar to a periodic heartbeat signal from the set of acceleration signal segments based on the heartbeat periodicity feature includes:

[0021] Each distance angle bin is sequentially used as a candidate interval, all the minimum points in each acceleration signal segment within each candidate interval are marked, and the acceleration signal segments within the preset time before and after each minimum point are intercepted. The intercepted acceleration signal segments are saved as samples in the template learning sample set;

[0022] The average value of the acceleration signal in the template learning sample set is used as the acceleration template, and the similarity between each sample in the template learning sample set and the acceleration template is calculated respectively, and the acceleration signal segment corresponding to the candidate interval with the highest similarity is located.

[0023] Optionally, the step of locating the acceleration signal segment corresponding to the candidate interval with the highest similarity further includes:

[0024] If the similarity is lower than the preset threshold, the preset time is changed, the template learning sample set is updated according to the changed preset time, the updated template learning sample set is used to recalculate the similarity between each sample and the acceleration template, and the acceleration signal segment is relocated.

[0025] Optionally, the step of segmenting the beat-by-beat heartbeat signal from the located acceleration signal segment includes:

[0026] Using a sliding time window with a step size of 1 sampling point and the same length as the acceleration template, the correlation coefficient function between the acceleration template in the acceleration signal segment and the acceleration signal in each sliding time window is calculated in sequence;

[0027] A plurality of positive peaks are obtained according to the calculated correlation coefficient functions;

[0028] The heartbeat interval sequence is calculated based on the timestamps between adjacent positive peaks;

[0029] The heartbeat interval sequence is averaged to obtain an average beat-by-beat heart rate, thereby obtaining a heart rate value.

[0030] Optionally, the step of calculating the heartbeat interval sequence based on the timestamps between adjacent positive peaks further includes:

[0031] Outliers in the heart beat interval sequence are eliminated based on physiological plausibility constraints.

[0032] In a second aspect, the present application also provides a heart rate monitoring system based on millimeter wave radar, which includes:

[0033] A phase signal extraction module is used to receive radar echo signals and extract the phase signal of the target monitoring object in the radar echo signals;

[0034] an acceleration signal extraction module, configured to process the phase signal using second-order differential to obtain a set of acceleration signal segments corresponding to the phase signal;

[0035] a heartbeat signal positioning module, configured to locate, from the set of acceleration signal segments, an acceleration signal segment that is most similar to a periodic heartbeat signal based on the periodic characteristics of the heartbeat;

[0036] The heart rate calculation module is used to segment the beat-by-beat heartbeat signal from the located acceleration signal segment, and calculate the heart rate based on the segmented beat-by-beat heartbeat signal.

[0037] In a third aspect, the present application provides a heart rate monitoring device, which includes: a memory and a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the heart rate monitoring method as described is implemented.

[0038] In a fourth aspect, the present application provides a computer storage medium, wherein a heart rate monitoring program is stored on the computer-readable storage medium, and when the heart rate monitoring program is executed by a processor, the steps of the heart rate monitoring method as described are implemented.

[0039] Beneficial effects:

[0040] The present invention provides a heart rate monitoring method, system, device, and storage medium based on millimeter-wave radar. The method comprises receiving a radar echo signal and extracting a phase signal of a target monitoring object from the radar echo signal; processing the phase signal using second-order differentials to obtain an acceleration signal segment corresponding to the phase signal; locating an acceleration signal segment that is most similar to a periodic heartbeat signal from a set of acceleration signal segments based on the periodic characteristics of the heartbeat; segmenting a beat-by-beat heartbeat signal from the located acceleration signal segment, and calculating the heart rate based on the segmented beat-by-beat heartbeat signal. The present invention locates the acceleration signal segment of the optimal distance angle bin for the heartbeat based on the periodicity of the heartbeat signal, and segmenting the beat-by-beat heartbeat signal from the acceleration signal segment to calculate an accurate heart rate value. This method effectively solves the problem of misjudgment of the spectrum peak detection method under the conditions of multiple people coexisting, respiratory harmonics, and body motion interference, improves the accuracy of heart rate calculation, and provides a highly robust solution for accurate heart rate monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a flowchart of the steps of the heart rate monitoring method based on millimeter wave radar of the present invention;

[0042] Figure 2 This is a schematic diagram of the principle structure of the heart rate monitoring method according to an embodiment of the present invention;

[0043] Figure 3 Schematic diagram of the principle of distance fast Fourier transform in an embodiment of the present invention;

[0044] Figure 4 This is a signal distance-time intensity diagram before static clutter removal in an embodiment of the present invention;

[0045] Figure 5 This is a signal distance-time intensity diagram after static clutter removal in an embodiment of the present invention;

[0046] Figure 6 Schematic diagram of the algorithm corresponding to the beamforming strategy in an embodiment of the present invention;

[0047] Figure 7 Schematic diagram of an experimental scenario of a beamforming strategy in an embodiment of the present invention;

[0048] Figure 8 A distance-angle graph generated by the MVDR beamforming strategy in an embodiment of the present invention;

[0049] Figure 9 2 is a phase signal curve diagram before unwinding in an embodiment of the present invention;

[0050] Figure 10 2 is a phase signal curve diagram after unwrapping in an embodiment of the present invention;

[0051] Figure 11 is a comparison diagram of the phase signal, the corresponding acceleration signal and the electrocardiogram signal in an embodiment of the present invention;

[0052] Figure 12 (a) is a schematic diagram of generating a heartbeat template in an embodiment of the present invention; (b) is a schematic diagram of an acceleration signal in an embodiment of the present invention; (c) is a schematic diagram of sliding the acceleration template onto an acceleration signal segment to calculate a correlation coefficient and estimate the IBI through peak detection in an embodiment of the present invention;

[0053] Figure 13 2 is a block diagram of the principle structure of the heart rate monitoring system in an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0055] When monitoring vital signs, breathing and heartbeats trigger regular micro-movements in the chest and abdomen. For example, normal breathing can cause a periodic fluctuation of approximately 5 mm on the body surface, while heartbeats produce only weak vibrations of approximately 0.1 mm. For most adults, the reference range for heart rate is 0.8-2.0 Hz, and respiratory rate is 0.1-0.5 Hz. A vital sign monitoring radar transmits electromagnetic wave signals to the human body through a transmitter. When the signals reach the surface of the body, the tiny displacements of the chest wall caused by breathing and heartbeats modulate the phase of the reflected echo, causing it to carry information about the periodic motion of breathing and heartbeats. After denoising, the spectral characteristics of the echo phase signals are analyzed using time-frequency analysis methods to separate key physiological parameters such as respiratory rate and heart rate.

[0056] Although millimeter-wave radar shows potential in the field of vital signs monitoring, there are currently some technical challenges.

[0057] First, multiple target signals are difficult to separate. If multiple targets are located at different distances from the radar, the radar echo signals from these targets can be separated by the range dimension. However, due to the radar's inherent range resolution, if multiple targets are within the same range unit, their radar echo signals overlap in the range dimension, making it impossible to effectively extract the radar echo signal of a single target.

[0058] Second, there is a lack of highly accurate heart rate measurement algorithms. Current heart rate measurement methods are primarily limited by four core challenges: signal weakness (chest vibration amplitude is only at the micron level and easily masked by environmental noise), dynamic interference coupling (the spectrum of respiratory harmonics 0.1-0.5Hz overlaps with the heart rate fundamental frequency 0.8-2Hz, resulting in aliasing), multi-target interference (signal crosstalk in multi-body scenarios causes effective echo attenuation), and time-varying characteristics (heart rate variability causes spectral instability). Therefore, accurately locating the heartbeat signal and filtering out the respiratory signal from the composite signal to extract a high-precision heartbeat signal is a challenge that needs to be addressed urgently.

[0059] Although prior art methods disclose using an integrated empirical mode decomposition algorithm to decompose the signal time series and isolate the target signal, followed by a discrete wavelet transform algorithm to further separate the respiratory and heartbeat signals, and finally using a multi-signal classification algorithm to estimate the heart rate, this method only monitors the heart rate of a single target. Furthermore, if the radar echo signal of the target is weak or noisy, significant errors may occur in the decomposition and frequency estimation of the radar echo signal. Therefore, accurate heart rate measurement is not possible.

[0060] To overcome the aforementioned challenges of existing technologies, the present invention discloses a heart rate monitoring method, system, device, and storage medium based on millimeter-wave radar. By optimizing signal processing, this method achieves robust heart rate monitoring, enabling long-term and accurate heart rate monitoring in everyday life. To address the core challenges of multi-target vital sign detection using millimeter-wave radar, key breakthroughs are being made in multi-target echo separation and high-precision heart rate estimation in complex environments.

[0061] The following further describes a method, system, device and storage medium for monitoring heart rate based on millimeter wave radar disclosed in this embodiment in conjunction with the accompanying drawings. Figure 1 As shown, this embodiment provides a flow chart of the steps of a heart rate monitoring method based on millimeter wave radar, as shown in FIG. Figure 2 As shown in the figure, the principle framework diagram corresponding to this method is provided. Figure 1 and Figure 2As shown, the method of the present invention mainly includes the following main parts: signal preprocessing, heartbeat signal extraction based on second-order differential, heartbeat signal positioning based on periodicity, and heartbeat signal segmentation based on template matching. Among them, signal preprocessing mainly includes multi-target signal separation, static clutter filtering (suppressing fixed reflection interference in the environment (such as walls, furniture), retaining dynamic target signals), phase signal advance and unwrapping (based on demodulation and phase unwrapping technology, extracting high-sensitivity phase signals caused by chest displacement from radar echoes). Heartbeat signal extraction mainly uses second-order differentials to enhance the heartbeat signal masked in the original phase signal, extract the acceleration characteristics caused by the heartbeat, and significantly suppress respiratory interference. Heartbeat signal positioning uses a distance angle bin positioning method driven by the periodic characteristics of the heartbeat to locate the spatial direction dominated by the mechanical movement of the heart. Heartbeat signal segmentation uses template matching to segment the heartbeat, calculate the beat-by-beat heart rate and heart rate, and provide a highly robust solution for non-contact heart rate monitoring in an interference environment.

[0062] Specifically, the heart rate monitoring method includes:

[0063] Step S1: Receive a radar echo signal and extract a phase signal of a target monitoring object from the radar echo signal.

[0064] In an FMCW radar system, the transmitted linear frequency modulated continuous wave (LFM) is mixed with the transmitted signal after reflection from the target, generating an intermediate frequency (IF) signal whose frequency is proportional to the target's distance. In this step, the received radar echo signal undergoes a range-dimensional Fast Fourier Transform (FFT), combined with a beamforming algorithm to estimate and separate the distances and angles of multiple target signals within the monitoring area. Because the heartbeat signal is weak and susceptible to interference from clutter and noise, this step also involves filtering out static clutter and extracting the phase signal of the target monitored object.

[0065] In detail, the steps of receiving a radar echo signal and extracting a phase signal of a target monitoring object in the radar echo signal specifically include:

[0066] Step S11: Perform a fast Fourier transform on each chirp cycle in the radar echo signal in a fast time to obtain a distance-time matrix, and obtain the distance information of the target monitoring object based on the distance-time matrix.

[0067] When receiving radar echo signals, they are sampled at fixed intervals to form a discrete sequence in the fast time dimension. To further determine the distance information of the target monitoring object, a fast Fourier transform is performed on each chirp cycle (a chirp cycle is the time required for a linear frequency modulation signal to complete a complete linear frequency change, where the frequency can change linearly from low to high (upper chirp) or from high to low (lower chirp)) in the fast time dimension to obtain distance information.

[0068] Range Fast Fourier Transform (Range-FFT) is the core algorithm for achieving target range resolution in FMCW radar. Specifically, the intermediate frequency signal is sampled at equal intervals within each chirp cycle to form a discrete sequence in the fast time dimension. In order to determine the distance information of the reflecting object, each chirp cycle is subjected to a fast Fourier transform in the fast time dimension to obtain a range-time matrix, which is called Range-FFT. Figure 3 As shown in the Range-FFT process diagram, the time domain signal is mapped to the frequency domain. The frequency values ​​corresponding to the spectrum peaks directly reflect the target delay, and the target distance is then calculated according to the distance formula. Range bins are the core concept of the Range-FFT output, with each frequency bin corresponding to a range bin. Range-FFT can separate reflections from monitored targets at different distances into different range bins. By extracting the phase of the range bin corresponding to the target distance, heartbeat and respiration amplitude information can be obtained. The energy peaks in different range bins indicate the target location. The range resolution is determined by the bandwidth. If the distance between two monitored targets exceeds the range resolution, their spectrum peaks can be clearly separated.

[0069] Step S12: Using the time average intensity of each pixel in the distance-time matrix as background noise, remove the static background noise signal from the distance-time matrix.

[0070] In the actual process of vital sign signal detection, the human target is in a micro-motion state. The target micro-motion echo signal and background clutter are superimposed and received by the radar. Some higher-power clutter will affect the observation and analysis of the target signal and even drown out the human micro-motion signal. Therefore, it is necessary to remove static noise interference before signal processing. After the distance-time matrix is ​​calculated in the above steps, the background noise signal can be removed based on the time intensity of each pixel in the distance-time matrix. Figure 4 and 5 They are the time-distance intensity diagrams corresponding to the signals before and after static noise removal, Figure 4 and Figure 5 After comparison, it can be found that after the static noise is removed, the image signal-to-noise ratio is significantly improved.

[0071] In specific implementation, the phasor mean cancellation method is used to remove static noise from the radar received signal. The time average intensity of each pixel in the distance-time matrix is ​​used as the background noise, and then the background noise signal is subtracted from the distance-time matrix. The formula for eliminating static noise signals is:

[0072]

[0073] In the above formula, i = 1, 2, …, M, n = 1, 2, …, N, y[i, n] represents the two-dimensional data matrix before mean cancellation, y′[i, n] represents the two-dimensional data matrix after mean cancellation, M is the number of range cells, N is the total number of signal cycles, and i is the index of the echo signal in the nth cycle.

[0074] Step S13: Calculate the covariance matrix of the target monitoring object's echo signal based on the distance information of the target monitoring object obtained by positioning, perform beamforming based on the covariance matrix, and separate multiple target signals.

[0075] After static noise signal removal is completed, there may be multiple monitoring targets within the monitoring area. In complex scenarios, when multiple monitoring targets are in the same distance unit, the echo signals completely overlap in the distance domain, resulting in the inability of single-dimensional distance detection to distinguish the number of targets and their individual characteristics. Therefore, target detection methods based on the distance dimension have limitations when processing multiple monitoring targets at the same distance. In this step, in order to distinguish different monitoring targets in the same distance dimension, it is proposed to introduce angular dimension information and achieve target separation through spatial signal processing.

[0076] Specifically, the array configuration of the MIMO radar can sense the path difference of the target's reflected signal, which is reflected as the phase difference between the receiving channels. Based on the phase difference, the target angle can be estimated, laying the foundation for vital sign extraction. In order to accurately locate the signal of the reflected target, digital beamforming is performed on all antenna units of each distance tap. Figure 6 As shown, the antennas of each channel use a weight vector at a specific angle to make the outputs superimposed in the same direction, pointing the antenna beam in a specific direction, thereby achieving the separation of the reflected echo signals of different monitoring targets.

[0077] In one implementation, the minimum variance distortionless response (MVDR) beamforming strategy is used to separate the signals of different monitoring targets. The formula for the weighting vector in this beamforming strategy is:

[0078]

[0079] Wherein, R is the covariance matrix of the received data, and a(θ) is the steering vector corresponding to the target signal incident from the direction of θ. The covariance matrix is ​​calculated from the intermediate frequency signal obtained in the above step S11.

[0080] Combine Figure 7 and Figure 8 As shown, take the monitoring scene with two target monitoring objects as an example, Figure 7 As shown in , target 1 is at an angle of 30 degrees north-west to the radar, target 2 is at an angle of north-east to the radar, and the distances between target 1 and target 2 and the radar are both 1m. Using the MVDR beamforming algorithm proposed in this step, we can get the following: Figure 8 The distance-angle diagram shown is used to locate the positions of target 1 and target 2, thereby separating the signal corresponding to target 1 and the signal corresponding to target 2.

[0081] Because traditional beamforming is limited by the mainlobe width, it leads to severe crosstalk between multiple targets at the same distance. This paper uses an adaptive MVDR beamforming algorithm to adaptively adjust the beam pattern to minimize the total output signal power (including interference and noise) while maintaining a constant gain (no distortion) in the target direction. This achieves super-resolution angle estimation and actively suppresses interference from non-target directions.

[0082] Step S14: locate the target monitoring object according to the different positions corresponding to the separated multiple target signals.

[0083] After multiple target signals are separated, the target monitoring object can be located according to the location of the target monitoring object.

[0084] Step S15: Perform arctangent demodulation on the baseband signal corresponding to the located target monitoring object to obtain a phase signal of the target monitoring object.

[0085] The amplitude of human breathing and heartbeat signals is small, and the resulting chest displacement is difficult to detect relative to the radar's distance change. However, their periodic micro-motion will cause sensitive changes in the phase of the intermediate frequency signal. Therefore, the time-frequency domain change information of the demodulated phase signal can be used to reflect the displacement changes of the human chest, thereby obtaining the results of breathing and heartbeat frequency estimation. In the radar system, the linear frequency modulated continuous wave signal is divided into two paths by a power divider: one path is radiated to the target area by the transmitting antenna, and the other path is generated by an orthogonal phase shifter to generate a local oscillator reference signal with a phase difference of 90°, which is called the I / Q path. After the target reflection signal is mixed with the local oscillator reference signal, an orthogonal baseband signal containing the target micro-motion information is output. In order to avoid the phase ambiguity problem caused by the limitation of phase angle information, it is necessary to perform inverse tangent demodulation on the received and derived I / Q baseband signals to obtain the phase signal. The phase signal can be calculated using the following formula:

[0086]

[0087] In computers, the angle range of the first and second quadrants is 0-π, and the angle range of the third and fourth quadrants is -π-0. During signal calculation, when the angle changes from 0 to 2π, the computer will first change the angle from 0 to π, and then change -π to 0, causing the phase of the signal to be at π. A jump of 2π occurs, forming a phase wrapping phenomenon, which is unwrapped by performing an undoing of the original phase by ±2π. Figure 9 and Figure 10 They are the phase signals before and after unwrapping, so it can be seen that after unwrapping, a clear phase signal can be obtained.

[0088] Before receiving the radar echo signal, in this embodiment, a MIMO radar planar array is used to implement two-dimensional beam scanning of the scene, generating a range angle matrix to accurately locate the target's chest movement.

[0089] Step S2: Process the phase signal using second-order differential to obtain a set of acceleration signal segments corresponding to the phase signal.

[0090] Breathing is typically slow and steady, while the heartbeat involves rapid muscle contraction. Therefore, the heartbeat's acceleration differs significantly from that of breathing. Therefore, extracting the acceleration signal rather than directly extracting the displacement signal can suppress breathing and amplify the heartbeat component. Since acceleration is the second-order derivative of displacement, the second-order derivative of phase can be numerically calculated to locate the acceleration signal corresponding to the heartbeat signal.

[0091] After the phase signal in the radar echo signal is extracted in the above step, this step uses a differentiator with improved noise suppression based on least squares smoothing to extract low-frequency acceleration while reducing high-frequency system noise.

[0092] In one implementation, the step of processing the phase signal using second-order differential to obtain a set of acceleration signal segments corresponding to the phase signal includes:

[0093] The acceleration value corresponding to each sampling point in the phase signal is calculated respectively by using a differentiator based on least squares smoothing to obtain an acceleration signal corresponding to each sampling point in the phase signal. Each acceleration signal constitutes an acceleration signal segment corresponding to the phase signal.

[0094] Specifically, a specific sample is first selected from the phase signal, and then three sampling points before and after the specific sample are selected as samples for calculation. The calculation formula is:

[0095]

[0096] Where s″ represents the second-order derivative at a specific time sample, s j represents a sample j distance away from a specific sample, and h represents the frame period between consecutive samples. S0 is a specific sample, and the three samples before and after S0 are selected for calculation. Each calculation of S0 is a sliding change. For example, if the phase signal corresponds to a sequence of 6000 sample points, S0 slides from the 4th sample point to the 5997th sample point during the calculation process. After calculation, a sequence containing (5997-4+1) points is obtained. This sequence is the acceleration signal segment.

[0097] like Figure 11 is a comparison diagram of the corresponding phase signal, acceleration signal and ECG signal. Figure 11 It can be seen that the acceleration signal calculated based on the phase signal has a high similarity with the heartbeat signal.

[0098] Step S3: Locate the acceleration signal segment that is most similar to the periodic heartbeat signal from the acceleration signal segment set based on the heartbeat periodicity feature.

[0099] Although the morphology of cardiac motion varies between subjects and locations, the morphology of cardiac motion in the same location and for the same subject is similar and periodic. After calculating multiple acceleration signal segments corresponding to the phase signal in step S2, the position of the heartbeat signal can be located from the multiple acceleration signal segments.

[0100] This study proposes an adaptive signal localization method based on periodic correlation and iterative template optimization. The core idea is to screen out the most periodically repetitive cardiac motion features by quantifying the morphological consistency between signal segments. The periodic correlation measurement model is based on the Pearson correlation coefficient.

[0101] In detail, this step specifically includes:

[0102] Step S31: Take each distance angle bin as a candidate interval in turn, mark all the minimum points in each acceleration signal in each candidate interval, and intercept the acceleration signal fragments within the preset time before and after each minimum point, and save the intercepted acceleration signal fragments as samples in the template learning sample set.

[0103] Since each acceleration signal segment in the calculated acceleration signal set corresponds to a distance bin and an angle bin, and the distance bins and angle bins corresponding to each acceleration signal segment are different, in this step, the distance and angle bins (that is, the acceleration signal segments corresponding to the positions of a distance and an angle positioning) are taken as candidate intervals in turn, and the acceleration signal segments are intercepted to obtain the corresponding template learning sample set.

[0104] In one implementation, taking the data for a candidate range angle bin (candidate bin) as an example, all minimum points in a single heartbeat signal are marked, and 0.5-second signal segments are captured before and after each minimum point. These signal segments constitute the template learning sample set S = {s1, s2, …, sv}.

[0105] Step S32: using the average value of the acceleration signal in the template learning sample set as the acceleration template, respectively calculating the similarity between each sample in the template learning sample set and the acceleration template, and locating the acceleration signal segment corresponding to the candidate interval with the highest similarity.

[0106] The mean of the sample set is taken as template T. To determine the signal with the best heartbeat periodicity in the candidate bin, in one implementation, the similarity between the sample set signal segments is quantified by calculating the Pearson correlation coefficient between the sample set and the template. Using the calculated Pearson correlation coefficient, the system selects the acceleration signal segment with the highest similarity between the sample set segments from multiple candidate bins. If the similarity exceeds a certain threshold (the threshold is 0.6), the template and acceleration signal segment are directly output.

[0107] Furthermore, if the similarity is lower than the preset threshold, the duration of the preset time is changed, the template learning sample set is updated according to the changed preset time, and the similarity between each sample and the acceleration template is recalculated using the updated template learning sample set to relocate the acceleration signal segment.

[0108] Because heartbeats vary from person to person, there may be weak correlation between acceleration signal segments segmented with a fixed 1-second time duration. Therefore, a template is regenerated and the acceleration signal is re-segmented. The prior assumption that any 1-second signal segment has at least one periodic pattern can be easily achieved. Using a template matching algorithm, template T is periodically aligned with the sequence, resulting in segmented signal segments S′ = {s′1, s′2, …, s′m} with true beats (this process is similar to the heartbeat signal segmentation described below). Template T is then re-updated to the mean of the segmented signal set, and the output acceleration signal is selected from the candidate interval with the highest similarity.

[0109] In this step, the candidate region with the highest similarity is selected to correspond to the location of the acceleration signal (the corresponding bin). Different candidate intervals correspond to different distance bins and angle bins. The acceleration signal corresponding to each candidate region may correspond to a signal representing breathing, some may represent heartbeat signals, and some may be interference signals that cannot represent vital signs. Therefore, in this step, a bin (a certain position and angle) with the optimal heartbeat is selected to calculate the heart rate. In other words, for a target monitoring object, first locate the bin with the highest energy. Then, with this bin as the center, define two distance bins in front and behind, and search within an angle range of 10° to the left and right. Within this range, there are many 60-second acceleration signal segments. By comparing the periodicity of these acceleration signals, the acceleration signal with the best periodicity is selected as the subsequent heart rate calculation signal.

[0110] Step S4: Segment the beat-by-beat heartbeat signal from the located acceleration signal segment, and calculate the heart rate based on the segmented beat-by-beat heartbeat signal.

[0111] In contactless vital sign monitoring, accurate segmentation of heartbeat signals is a key step in achieving heart rate estimation. Although individual heartbeat amplitudes vary across subjects, or within the same subject, the waveforms of heartbeats at the same location exhibit significant temporal-shape similarities. Based on this characteristic, after locating acceleration signal segments, a template matching algorithm proposed in this paper performs sliding correlation analysis on these segments to achieve heartbeat signal segmentation and heart rate calculation.

[0112] Specifically, the step of segmenting the beat-by-beat heartbeat signal from the located acceleration signal segment includes:

[0113] Step S41: Using a sliding time window with a step size of 1 sampling point and the same length as the acceleration template, sequentially calculate the correlation coefficient function between the acceleration template in the target acceleration signal segment and the acceleration signal in each sliding time window.

[0114] Template Construction and Sliding Correlation Matching: The template selected is the template output in the previous step, and the matching signal is the corresponding acceleration signal. Using a sliding time window with the same length as the template and a step size of 1 sampling point, calculate the correlation coefficient function between the template and the heartbeat signal within the sliding time window.

[0115] Step S42: Obtain multiple positive peaks according to the calculated correlation coefficient functions.

[0116] The more similar the heartbeat signal within the time window is to the template, the higher the correlation coefficient. When the heartbeat signal within the time window is highly similar to the template, a positive peak will always appear in the correlation coefficient function. Detecting a significant positive peak in the correlation coefficient sequence corresponds to the onset of the heartbeat.

[0117] Step S43: Calculate a heartbeat interval sequence based on the timestamps between adjacent positive peaks; average the heartbeat interval sequence to obtain an average beat-by-beat heart rate, and obtain a heart rate value.

[0118] Combine Figure 12 The following table shows the generated heartbeat module, acceleration signal, and calculated correlation coefficient. Based on the timestamps of adjacent peaks, a heartbeat interval sequence (i.e., an IBI sequence) is calculated. Based on physiological plausibility constraints, outliers are removed to construct a valid IBI sequence. The valid IBI sequence is averaged to obtain the beat-by-beat average heart rate, which is then used to determine the heart rate value.

[0119] This step proposes a template matching method that uses learnable waveform priors and correlation analysis to accurately estimate heartbeat times. This method helps further evaluate heart rate metrics and reduces abnormally large heart rate estimation errors that can be caused by incorrect frequency estimation. The proposed algorithm effectively addresses the misjudgment issues of spectrum peak detection methods in the presence of multiple people, respiratory harmonics, and body motion interference, providing a highly robust solution for the practical application of radar vital sign monitoring technology.

[0120] On the basis of the above-mentioned heart rate monitoring method, the present invention also provides a heart rate monitoring system based on millimeter wave radar. Figure 13 As shown, including:

[0121] The phase signal extraction module 1301 is used to receive the radar echo signal and extract the phase signal of the target monitoring object in the radar echo signal; its function is as described in step S1.

[0122] The acceleration signal extraction module 1302 is used to process the phase signal using second-order differential to obtain a set of acceleration signal segments corresponding to the phase signal; its function is as described in step S2.

[0123] The heartbeat signal positioning module 1303 is used to locate the acceleration signal segment that is most similar to the periodic heartbeat signal from the acceleration signal segment set based on the periodic characteristics of the heartbeat; its function is as described in step S3.

[0124] The heart rate calculation module 1304 is used to segment the beat-by-beat heartbeat signal from the located acceleration signal segment and calculate the heart rate based on the segmented beat-by-beat heartbeat signal. Its function is as described in step S4.

[0125] The present application provides a heart rate monitoring device, which includes: a memory and a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the heart rate monitoring method as described above is implemented.

[0126] The present application provides a computer storage medium, wherein a heart rate monitoring program is stored on the computer-readable storage medium. When the heart rate monitoring program is executed by a processor, the steps of the heart rate monitoring method as described above are implemented.

[0127] In order to verify the heart rate monitoring effect of the method provided by the present invention in a specific application, relevant verification is carried out through experimental verification below.

[0128] In application, two Tx antennas and four Rx antennas were configured in time-division multiplexing transmission mode. Data acquisition was performed in a static environment. During experimental acquisition, interference from other objects, including tables and chairs, was minimized. The target subject breathed normally, sat upright, with their chest facing the radar sensor, and remained quasi-static to minimize interference from limb movement. To obtain a realistic heartbeat signal, a multi-parameter acquisition device simultaneously collected the target subject's ECG and respiration signals while the radar collected their respiration and heartbeat signals. The multi-parameter acquisition device used a three-lead ECG mode: the positive and negative electrodes were attached to the left and right wrists, respectively, and the ground electrode was connected to the right ankle. A breathing belt was tied around the chest, and the sampling rate was uniformly set to 1000Hz. Data was collected for 60 seconds for each target subject, with data repeated three times at each distance point. In a single-target scenario, the radar was fixed directly in front of the target subject using a bracket. Multiple scenario validation was conducted at three detection distances: 0.5m, 1.0m, and 1.5m. In a multi-target scenario, two target monitoring objects sit quietly in the radar detection area at azimuth angles of 30° and -30°, respectively. The distance parameters are set to 1.0m and 1.5m, respectively, forming a monitoring scenario with the same distance at different angles. The heart rate estimation error data of the target monitoring object on the left side of the radar (30° azimuth) at different distances are extracted.

[0129] Table 1: Heart rate estimation error of the target monitored object at different distances

[0130]

[0131] The experimental data of this study fully verified the clinical-grade detection accuracy and scenario adaptability of the proposed algorithm. As shown in Table 1, in the single-target monitoring scenario (no inter-subject interference), the system achieved medical-grade detection accuracy (mean absolute error ≤ 2bpm) within the detection distance range of 0.5-1.5 meters. Specifically, at a close distance of 0.5 meters, the heart rate error was only 0.65bpm, and the error was further reduced to 0.59bpm at a medium distance of 1 meter. Even when extended to a long-range scenario of 1.5 meters, the error was still strictly controlled to 1.03bpm through the time domain periodic stability constraint. In the multi-target separation scenario, the algorithm demonstrated a breakthrough spatial resolution capability: when the two subjects were at a distance of 1 meter and 1.5 meters respectively, the system separated the multi-target signals through spatial filtering and achieved accurate heart rate calculation. The detection errors were 1.93bpm and 0.88bpm, respectively. This performance breakthrough stems from two innovative mechanisms: 1) a periodic intelligent bin selection algorithm that locks on to the optimal signal source to improve the target signal-to-noise ratio; 2) a dynamic template matching mechanism that uses the Pearson correlation coefficient to achieve precise segmentation of heartbeat waveforms, effectively suppressing signal aliasing caused by multipath reflections.

[0132] Experimental results show that the system can accurately locate the target's heart movement and achieve robust heart rate measurement. The proposed MIMO radar algorithm framework has potential applications in the fields of emotion recognition and sleep quality assessment.

[0133] The present invention provides a heart rate monitoring method, system, device, and storage medium based on millimeter-wave radar. The method comprises receiving a radar echo signal and extracting a phase signal of a target monitoring object from the radar echo signal; processing the phase signal using second-order differentials to obtain an acceleration signal segment corresponding to the phase signal; locating an acceleration signal segment that is most similar to a periodic heartbeat signal from the acceleration signal segment corresponding to the phase signal based on the periodic characteristics of the heartbeat; segmenting a beat-by-beat heartbeat signal from the located acceleration signal segment, and calculating the heart rate based on the segmented beat-by-beat heartbeat signal. The present invention locates an acceleration signal segment of an optimal distance angle bin for the heartbeat based on the periodicity of the heartbeat signal, and segmenting a beat-by-beat heartbeat signal from the acceleration signal segment, thereby calculating an accurate heart rate value. This method effectively solves the problem of misjudgment of a spectrum peak detection method under the conditions of multiple people coexisting, respiratory harmonics, and body motion interference, improves the accuracy of heart rate calculation, and provides a highly robust solution for accurate heart rate monitoring.

[0134] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or N embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0135] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. A heart rate monitoring method based on millimeter wave radar, characterized in that: include: Receive radar echo signals and extract the phase signal of the target monitoring object in the radar echo signals; Processing the phase signal using second-order differential to obtain a set of acceleration signal segments corresponding to the phase signal; Locating, from the set of acceleration signal segments, an acceleration signal segment that is most similar to a periodic heartbeat signal based on the heartbeat periodicity feature; A beat-by-beat heartbeat signal is segmented from the located acceleration signal segment, and the heart rate is calculated based on the segmented beat-by-beat heartbeat signal.

2. The heart rate monitoring method based on millimeter wave radar according to claim 1, characterized in that: The step of receiving the radar echo signal and extracting the phase signal of the target monitoring object in the radar echo signal comprises: Perform fast Fourier transform on each chirp cycle in the radar echo signal in a fast time to obtain the distance-time matrix. Based on the distance-time matrix, the distance information of the target monitoring object is obtained. The time average intensity of each pixel in the distance-time matrix is ​​used as the background noise to remove the static background noise signal from the distance-time matrix; Calculate the covariance matrix of the target monitoring object's echo signal based on the distance information of the target monitoring object obtained by positioning, perform beamforming based on the covariance matrix, and separate multiple target signals; The target monitoring object is located according to the different positions corresponding to the multiple separated target signals; Perform arc tangent demodulation on the baseband signal corresponding to the located target monitoring object to obtain the phase signal of the target monitoring object.

3. The heart rate monitoring method based on millimeter wave radar according to claim 2, characterized in that: The step of processing the phase signal by using second-order differential to obtain a set of acceleration signal segments corresponding to the phase signal includes: The acceleration value corresponding to each sampling point in the phase signal is calculated respectively by using a differentiator based on least squares smoothing to obtain an acceleration signal corresponding to each sampling point in the phase signal. Each acceleration signal constitutes an acceleration signal segment set corresponding to the phase signal.

4. The heart rate monitoring method based on millimeter wave radar according to claim 1, characterized in that: The step of locating the acceleration signal segment that is most similar to the periodic heartbeat signal from the acceleration signal segment set based on the heartbeat periodicity feature includes: Each distance angle bin is sequentially used as a candidate interval, all the minimum points in each acceleration signal segment within each candidate interval are marked, and the acceleration signal segments within the preset time before and after each minimum point are intercepted. The intercepted acceleration signal segments are saved as samples in the template learning sample set; The average value of the acceleration signal in the template learning sample set is used as the acceleration template, and the similarity between each sample in the template learning sample set and the acceleration template is calculated respectively, and the acceleration signal segment corresponding to the candidate interval with the highest similarity is located.

5. The heart rate monitoring method based on millimeter wave radar according to claim 4, characterized in that: The step of locating the acceleration signal segment corresponding to the candidate interval with the highest similarity also includes: If the similarity is lower than the preset threshold, the preset time is changed, the template learning sample set is updated according to the changed preset time, the updated template learning sample set is used to recalculate the similarity between each sample and the acceleration template, and the acceleration signal segment is relocated.

6. The heart rate monitoring method based on millimeter wave radar according to claim 4, characterized in that: The step of segmenting the beat-by-beat heartbeat signal from the located acceleration signal segment comprises: Using a sliding time window with a step size of 1 sampling point and the same length as the acceleration template, the correlation coefficient function between the acceleration template in the acceleration signal segment and the acceleration signal in each sliding time window is calculated in sequence; A plurality of positive peaks are obtained according to the calculated correlation coefficient functions; The heartbeat interval sequence is calculated based on the timestamps between adjacent positive peaks; The heartbeat interval sequence is averaged to obtain an average beat-by-beat heart rate, thereby obtaining a heart rate value.

7. The heart rate monitoring method based on millimeter wave radar according to claim 6, characterized in that: The step of calculating the heartbeat interval sequence according to the timestamps between adjacent positive peaks further includes: Outliers in the heart beat interval sequence are eliminated based on physiological plausibility constraints.

8. A heart rate monitoring system based on millimeter wave radar, characterized in that: include: A phase signal extraction module is used to receive radar echo signals and extract the phase signal of the target monitoring object in the radar echo signals; an acceleration signal extraction module, configured to process the phase signal using second-order differential to obtain a set of acceleration signal segments corresponding to the phase signal; a heartbeat signal positioning module, configured to locate, from the set of acceleration signal segments, an acceleration signal segment that is most similar to a periodic heartbeat signal based on the periodic characteristics of the heartbeat; The heart rate calculation module is used to segment the beat-by-beat heartbeat signal from the located acceleration signal segment and calculate the heart rate based on the segmented beat-by-beat heartbeat signal.

9. A heart rate monitoring device, characterized in that: include: A memory and a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the heart rate monitoring method according to any one of claims 1 to 7 is implemented.

10. A computer storage medium, characterized in that The computer-readable storage medium stores a heart rate monitoring program, and when the heart rate monitoring program is executed by the processor, the steps of the heart rate monitoring method according to any one of claims 1 to 7 are implemented.

Images