A scene weak constraint heartbeat signal extraction method based on millimeter wave perception

Abstract

The application relates to the technical field of radar signal processing and vital sign monitoring, in particular to a scene weak constraint heartbeat signal extraction method based on millimeter wave perception. Firstly, a micro-motion frequency band energy accumulation and adaptive beam forming algorithm are used to realize robust positioning and spatial signal enhancement of a human target; secondly, a weighted fusion model based on multi-distance gate phase consistency is constructed, and a circle fitting and differential cross-multiplication algorithm is combined to reconstruct a high-quality phase signal under a low signal-to-noise ratio; finally, a cascaded frequency domain harmonic trap and a physical enhancement algorithm based on a weighted noise suppression differentiator are used to eliminate respiratory interference and highlight the heart beat feature. The method can effectively solve the technical problems of heartbeat signal extraction difficulty and waveform distortion in a complex environment, and further realizes reconstruction of a high-fidelity heartbeat waveform.

Description

Technical Field

[0001] The embodiments of this application relate to the fields of radar signal processing and vital sign monitoring technology, and in particular to a method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing. Background Technology

[0002] With the rapid development of smart healthcare and wireless sensing technologies, non-contact vital sign monitoring technology based on millimeter-wave radar has attracted much attention. Compared to contact sensors, radar monitoring has significant advantages such as being unconstrained and offering better privacy protection. However, existing high-precision radar heartbeat monitoring research typically requires the human body to be at an extremely close distance to the radar and to strictly maintain a direct orientation towards the radar antenna. While this highly constrained monitoring mode can achieve a high signal-to-noise ratio, it severely limits the user's posture and range of motion in practical applications, making it difficult to meet the needs of seamless and continuous monitoring in daily life.

[0003] To overcome this limitation, related technologies focus on vital sign monitoring in weakly constrained scenarios, aiming to achieve accurate heartbeat monitoring even when the human body is not directly facing the radar or is far from the radar. However, this transition from strong to weakly constrained scenarios presents several severe challenges to signal processing: First, long distances cause a sharp drop in the signal-to-noise ratio of the radar echo signal; second, complex indoor environments cause multipath effects, resulting in drastic fluctuations in the human body's radar cross-section; and finally, the amplitude of respiratory movements is much greater than that of the heartbeat, and its higher harmonics easily cover the heartbeat frequency band.

[0004] Traditional signal processing methods typically rely on the maximum energy value in a single frame for localization and employ simple bandpass filtering or time-domain differential extraction to extract the heartbeat signal. However, under weak constraints, single-distance gate signals are prone to signal loss due to slight movements of the human body, and linear filters cannot effectively eliminate respiratory harmonic interference in the same frequency band. This results in distorted waveforms and blurred features in the extracted heartbeat signal, failing to meet the needs of subsequent fine-grained cardiac state analysis. Summary of the Invention

[0005] In view of this, embodiments of this application propose a method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing. This method can effectively solve the technical problems of difficult heartbeat signal extraction and waveform distortion in complex environments, thereby achieving high-fidelity reconstruction of heartbeat waveforms.

[0006] To achieve the above objectives, embodiments of this application propose a method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing. The method includes the following steps: Millimeter-wave radar is used to collect echo signals containing human targets, and the echo signals are subjected to range-dimensional fast Fourier transform to obtain the range-slow time matrix. Based on the range-slow time matrix, energy accumulation is performed within the micro-motion frequency band to determine the target range gate where the human target is located, and an adaptive beamforming algorithm is used at the target range gate to perform spatial filtering and enhancement on the multi-channel signal; Centered on the target range gate, multiple consecutive range gates are selected to form a region of interest. At the same time, static clutter suppression and phase demodulation are performed on the signal of each range gate within the region of interest to obtain the phase sequence of each range gate. Weighted fusion based on phase consistency measurement between phase sequences yields enhanced vital sign signals. The enhanced vital signs signal was subjected to spectral analysis, and the respiratory fundamental frequency was adaptively estimated. At the same time, a cascaded harmonic notch filter was used to filter out the respiratory fundamental frequency and its harmonic components, resulting in a signal after filtering out respiratory interference. A noise suppression differentiator based on the weighted least squares principle is used to differentiate the signal after filtering out respiratory interference, and the displacement signal is converted into an acceleration signal to output a high-fidelity heartbeat signal.

[0007] To achieve the above objectives, embodiments of this application also propose a scene weakly constrained heartbeat signal extraction system based on millimeter-wave sensing, the system comprising: The signal acquisition and preprocessing module is used to acquire echo signals containing human targets using millimeter-wave radar, and to perform range-dimensional fast Fourier transform on the echo signals to obtain the range-slow time matrix. The target localization and filtering enhancement module is used to accumulate energy within the micro-motion frequency band based on the range-slow time matrix to determine the target range gate where the human target is located, and to use an adaptive beamforming algorithm to perform spatial filtering and enhancement on the multi-channel signal at the target range gate. The clutter suppression and phase demodulation module is used to select multiple consecutive range gates to form a region of interest centered on the target range gate, and simultaneously perform static clutter suppression and phase demodulation on the signal of each range gate within the region of interest to obtain the phase sequence of each range gate; The phase consistency weighted fusion module is used to perform weighted fusion based on the phase consistency measurement between each phase sequence to obtain enhanced vital sign signals. The respiratory harmonic adaptive notch filter module is used to perform spectral analysis on the enhanced vital signs signal and adaptively estimate the respiratory fundamental frequency. At the same time, it uses a cascaded harmonic notch filter to filter out the respiratory fundamental frequency and its harmonic components, thus obtaining a signal after filtering out respiratory interference. The signal output module is used to differentiate the signal after filtering out breathing interference using a noise suppression differentiator based on the weighted least squares principle, and convert the displacement signal into an acceleration signal to output a high-fidelity heartbeat signal.

[0008] To achieve the above objectives, embodiments of this application also propose an electronic device, including: a processor and a memory, wherein the memory stores instructions executable by the processor, and the processor is configured to execute the instructions such that the electronic device can implement the scene weak constraint heartbeat signal extraction method based on millimeter wave sensing as described above.

[0009] To achieve the above objectives, embodiments of this application also propose a computer-readable storage medium storing a computer program that, when executed by a processor, enables the implementation of a scene weak constraint heartbeat signal extraction method based on millimeter-wave sensing as described above.

[0010] This application proposes a method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing. First, millimeter-wave radar is used to acquire echo signals containing human targets. A range-dimensional Fast Fourier Transform (FFT) is performed on the echo signals to obtain a range-slow-time matrix, enabling non-contact detection of distant human targets and converting the signal to a dimension capable of distinguishing targets at different distances. Then, based on the range-slow-time matrix, energy accumulation is performed within a micro-motion frequency band to determine the target range gate where the human target is located. An adaptive beamforming algorithm is then used at the target range gate to perform spatial filtering and enhancement on the multi-channel signals. This improves the robustness and accuracy of target range gate detection through energy accumulation within the frequency band, avoiding positioning failures caused by single-frame signal fluctuations. Simultaneously, adaptive beamforming achieves spatial focusing in the target direction, significantly improving the echo signal-to-noise ratio. Finally, multiple consecutive range gates are selected as the region of interest (ROI) centered on the target range gate. Static clutter suppression and phase demodulation are performed on the signals of each range gate within the ROI to obtain the range gates for each target range gate. The phase sequence of the distance gate is weighted and fused based on the phase consistency metric between the phase sequences to obtain an enhanced vital sign signal. This method utilizes the extended target characteristics of the human chest cavity to transform a single signal source into multiple signal sources, achieving spatial diversity. Furthermore, phase consistency weighted fusion yields a stable vital sign signal. Next, spectral analysis is performed on the enhanced vital sign signal, and the respiratory fundamental frequency is adaptively estimated. Simultaneously, a cascaded harmonic notch filter is used to filter out the respiratory fundamental frequency and its harmonic components. This ensures accurate estimation of the respiratory fundamental frequency, avoiding incomplete filtering or signal damage caused by estimation errors. Filtering out the respiratory fundamental frequency and its main harmonics also prevents respiratory interference with the heartbeat frequency band. Finally, a noise suppression differentiator based on the weighted least squares principle is used to differentiate the signal after filtering out respiratory interference, converting the displacement signal into an acceleration signal, and outputting a high-fidelity heartbeat signal. Based on this, the proposed solution effectively addresses the technical problems of difficult heartbeat signal extraction and waveform distortion in complex environments, thereby achieving high-fidelity heartbeat waveform reconstruction.

[0011] Optionally, the adaptive beamforming algorithm is the Capon beamforming algorithm; the optimal weight vector of the Capon beamforming algorithm... The calculation formula is: ; in, The covariance matrix of the received signal. For scanning angle The corresponding guide vector, express The conjugate transpose of .

[0012] Optionally, static clutter suppression employs a circle fitting algorithm, which estimates the center of the circle by minimizing the sum of squared algebraic distances from data points to the fitted circle. Its linear least squares solution is: ; ; in, The design matrix is ​​constructed based on the real and imaginary parts of the signal. The observation vector is constructed based on the real and imaginary parts of the signal; Indicates auxiliary parameters, Given the radius of the circle, estimate the center of the circle. This is the static clutter vector.

[0013] Optionally, phase demodulation employs a differential and cross-multiplication algorithm, with its discrete phase increment... The calculation formula is: ; in, and They are time points The real and imaginary parts of the denoised signal.

[0014] Optionally, weighted fusion is performed based on the consistency measure between each phase sequence to obtain enhanced vital sign signals, including: Calculate the median reference sequence for all range-gated phase sequences; The normalized cross-correlation coefficients of each phase sequence and the median reference sequence are used as a phase consistency measure. The fusion weights are calculated based on the phase consistency measure, and the multi-range gate phase signals are weighted and summed to obtain the enhanced vital signs signal.

[0015] Optionally, the median reference sequence of all distance gate phase sequences is denoted as , ;in, Indicates the total number of doors at all distances; The normalized cross-correlation coefficient between each phase sequence and the median reference sequence is used as a measure of phase consistency, and its calculation formula is as follows: No. Consistency measure of distance gates for: ; in, Indicates the length of the processing window. For the first A distance gate at time phase sequence, For this phase sequence in the window The mean within, Median reference sequence In the window The mean within; Assigned to the Fusion weights of phase sequences of distance gates The calculation formula is as follows: ; in, Indicates all Phase consistency metric of distance gates Perform summation; The weighted summation of the multi-range gate phase signals yields an enhanced vital sign signal. The calculation formula is as follows: .

[0016] Optionally, adaptive estimation of respiratory fundamental frequency A two-stage strategy is adopted: In the first stage, the maximum value of the spectrum within the preset respiratory rate range is searched to obtain a coarse estimate of the frequency. ; In the second stage, the frequency that maximizes the cumulative energy of the fundamental frequency and its harmonics within the neighborhood of the coarsely estimated frequency is searched as the finely estimated respiratory fundamental frequency. Among them, accumulated energy The calculation formula is as follows: ; in, Indicating enhanced vital signs The spectrum obtained after Fast Fourier Transform Indicates the harmonic order. This represents the candidate respiratory fundamental frequencies searched by traversing the neighborhood of the coarse estimate. This represents the integral half-bandwidth of the spectrum.

[0017] Optionally, a noise suppression differentiator based on the weighted least squares principle performs differentiation processing on the signal after filtering out breathing interference, and its second-order differential estimation... The formula is: ; in, Indicates the filter length. Indicates the center of symmetry. Indicates the sampling time interval of the signal. This represents the weighting coefficient for the current calculation point. The signal value at the current calculation point. For the first The weighting coefficients of the symmetrical sampling points, and These represent the points relative to the current calculation point, i.e., the i-th... The first and forward The signal value at each sampling point. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies of this application will be briefly introduced below. Obviously, the following drawings are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. The drawings described herein are only used to explain this application and are not intended to limit this application.

[0019] Figure 1 This is a flowchart of a method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing, provided in one embodiment of this application; Figure 2 This is a comparison diagram of the effects of a related technology provided in one embodiment of this application and the adaptive beamforming angle estimation provided in an embodiment of this application; Figure 3 This is a comparison diagram of the effects of different phase extraction algorithms under large respiratory movements provided in one embodiment of this application; Figure 4 This is a schematic diagram of a scene weak constraint heartbeat signal extraction system based on millimeter wave sensing provided in another embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in another embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the various embodiments of this application will be described in detail below with reference to the accompanying drawings. Those skilled in the art will understand that many technical details have been presented in the embodiments of this application to facilitate better understanding. However, the technical solutions claimed in this application can be implemented even without these technical details and various variations and modifications based on the following embodiments. The division of the following embodiments is for ease of description and should not constitute any limitation on the specific implementation of this application. The following embodiments can be combined with and referenced by each other without contradiction.

[0021] With the rapid development of smart healthcare and wireless sensing technologies, non-contact vital sign monitoring technology based on millimeter-wave radar has attracted much attention. Compared to contact sensors, radar monitoring has significant advantages such as being unconstrained and offering better privacy protection. However, existing high-precision radar heartbeat monitoring research typically requires the human body to be at an extremely close distance to the radar and to strictly maintain a direct orientation towards the radar antenna. While this highly constrained monitoring mode can achieve a high signal-to-noise ratio, it severely limits the user's posture and range of motion in practical applications, making it difficult to meet the needs of seamless and continuous monitoring in daily life.

[0022] To overcome this limitation, related technologies focus on vital sign monitoring in weakly constrained scenarios, aiming to achieve accurate heartbeat monitoring even when the human body is not directly facing the radar or is far from the radar. However, this transition from strong to weakly constrained scenarios presents several severe challenges to signal processing: First, long distances cause a sharp drop in the signal-to-noise ratio of the radar echo signal; second, complex indoor environments cause multipath effects, resulting in drastic fluctuations in the human body's radar cross-section; and finally, the amplitude of respiratory movements is much greater than that of the heartbeat, and its higher harmonics easily cover the heartbeat frequency band.

[0023] Traditional signal processing methods typically rely on the maximum energy value in a single frame for localization and employ simple bandpass filtering or time-domain differential extraction to extract the heartbeat signal. However, under weak constraints, single-distance gate signals are prone to signal loss due to slight movements of the human body, and linear filters cannot effectively eliminate respiratory harmonic interference in the same frequency band. This results in distorted waveforms and blurred features in the extracted heartbeat signal, failing to meet the needs of subsequent fine-grained cardiac state analysis.

[0024] In view of this, embodiments of this application propose a method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing. This method can effectively solve the technical problems of difficult heartbeat signal extraction and waveform distortion in complex environments, thereby achieving high-fidelity reconstruction of heartbeat waveforms.

[0025] For example, the method proposed in the embodiments of this application can be used for weakly constrained scenarios such as long distance, non-direct angle, and strong breathing interference in non-contact monitoring.

[0026] The embodiments of this application also propose an adaptive space-time-frequency domain enhancement architecture. First, robust localization and spatial signal enhancement of human targets are achieved using micro-motion frequency band energy accumulation and the Capon beamforming algorithm. Second, a weighted fusion model based on multi-range gate phase consistency is constructed, combined with circle fitting and differential cross-multiplication algorithms, to reconstruct high-quality phase signals under low signal-to-noise ratio conditions. Finally, respiratory interference is eliminated and cardiac pulsation characteristics are highlighted through a cascaded frequency domain harmonic notch filter and a physical enhancement algorithm based on a weighted noise suppression differentiator. Therefore, the embodiments of this application can effectively solve the technical problems of difficult heartbeat signal extraction and waveform distortion in complex environments, achieving high-fidelity reconstruction of heartbeat waveforms.

[0027] For a detailed description of the methods proposed in the embodiments of this application, please refer to the following embodiments.

[0028] One embodiment of this application proposes a method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing, applied to an electronic device. The electronic device can be a terminal or a server; this embodiment and subsequent embodiments will use a server as an example. The implementation details of the method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing proposed in this embodiment are described below. The following implementation details are provided for ease of understanding and are not essential for implementing this solution.

[0029] The specific process of the scene weak constraint heartbeat signal extraction method based on millimeter wave sensing proposed in this embodiment can be described as follows: Figure 1 As shown, it includes: Step 101: Use millimeter-wave radar to collect echo signals containing human targets, and perform a range-dimensional fast Fourier transform on the echo signals to obtain a range-slow time matrix.

[0030] For example, the distance-slow-time matrix contains information about the micro-motions of the human target in the distance dimension. For example, chest displacement caused by breathing and heartbeat.

[0031] For example, millimeter-wave radar can be deployed in a monitoring area to collect echo signals containing human targets. Millimeter-wave radar can use a 60 GHz frequency for radar data acquisition. First, a frequency-modulated continuous wave (FMCW) signal is transmitted. When the signal encounters a human target, the echo signal is received by the radar. First, the raw echo signal is preprocessed, including mixing and analog-to-digital conversion, to obtain an intermediate frequency (IF) signal. Then, a range-fast Fourier transform (Range-FFT) is performed on the radar signal from each antenna channel to obtain a range-slow time matrix. The rows of the matrix correspond to range gates (e.g., each range gate represents a specific range interval), and the columns correspond to slow time (e.g., the sampling time point of the radar frame).

[0032] Step 102: Based on the distance-slow time matrix, energy accumulation is performed in the micro-motion frequency band to determine the target distance gate where the human target is located, and an adaptive beamforming algorithm is used at the target distance gate to perform spatial filtering and enhancement on the multi-channel signal.

[0033] For example, an adaptive beamforming algorithm could be the Capon beamforming algorithm, also known as the Minimum Variance Distortionless Response (MVDR) algorithm. MVDR is an adaptive beamforming technique designed to minimize array output power while maintaining a distortion-free response to signals in the desired direction.

[0034] For example, to address the instability of instantaneous energy detection, a micro-motion frequency band can be defined, and the cumulative energy spectrum of multi-channel signals within this band can be calculated. Then, by utilizing the sparsity and periodicity of vital signs signals in the frequency domain, noise is smoothed and the target is highlighted through energy accumulation. Combined with constant false alarm rate detection, the center distance gate of the human target can be robustly locked.

[0035] For example, in obtaining the range-slow-time matrix, a micro-motion band is first defined to cover respiratory and heart rate frequencies. The micro-motion band is typically 0.1Hz to 5Hz. Spectral analysis is performed on the slow-time signal of each range gate to calculate the energy value within that band. Then, the energy of all antenna channels is incoherently accumulated to obtain the accumulated energy spectrum for each range gate. Next, a constant false alarm rate (CFAR) detection algorithm is used to search for peaks in the accumulated energy spectrum to determine the target range gate for the human target. .

[0036] Next, at the target distance from the door Spatial enhancement is performed using multi-antenna array data (e.g., 8 virtual channels). Then, the Capon adaptive beamforming algorithm is used to calculate the optimal weight vector. Specifically, the covariance matrix R of the received signal is first calculated, and then the steering vector for the scanning angle is constructed. The scanning range of the guide vector is typically -60° to 60°. Finally, the optimal weights are calculated using a formula, and a single-channel signal is synthesized.

[0037] In one possible implementation, the optimal weight vector of the Capon beamforming algorithm The calculation formula is: ; in, The covariance matrix of the received signal. For scanning angle The corresponding guide vector, express The conjugate transpose of .

[0038] It is understandable that the optimal weight vector While ensuring the gain in the target direction, the power in the interference direction is minimized. Finally, the multi-channel signals are weighted and fused to obtain the enhanced signal after spatial filtering, which can significantly improve the signal-to-noise ratio and the accuracy of target orientation estimation.

[0039] For example, at a defined range gate, the covariance matrix can be calculated using multi-antenna array data. The Capon beamforming algorithm is then used to calculate the optimal weight vector, minimizing interference power in other directions while maintaining target directional gain. This significantly improves the echo signal-to-noise ratio while achieving high-resolution target azimuth estimation.

[0040] Understandably, to verify the advantages of Capon's adaptive beamforming algorithm over conventional beamforming (CBF) in terms of angle estimation accuracy and spatial resolution, a set of dual-signal-source resolution simulation experiments can be designed. This experiment simulates a uniform linear array with an element spacing of half a wavelength. To test the algorithm's resolution capability with finite angular spacing, two signal sources are set up at azimuth angles of 0° and 20° respectively. To comprehensively evaluate the algorithm's performance under different noise levels, spatial spectrum estimation is performed under two typical noise environments: high signal-to-noise ratio (SNR) and low SNR.

[0041] Experimental results are as follows Figure 2 As shown, Figure 2 This is a comparison diagram of the effects of a related technology provided in one embodiment of this application and the adaptive beamforming angle estimation provided in an embodiment of this application; Figure 2(a) in the figure is used to represent the angle estimation effect under high signal-to-noise ratio environment; Figure 2 (b) in the figure is used to represent the angle estimation effect in a low signal-to-noise ratio environment.

[0042] from Figure 2 The significant differences in waveform morphology between the two algorithms are clearly observable: First, in high signal-to-noise ratio (SNR) environments, both algorithms can detect two targets located at 0° and 20°. However, limited by the Rayleigh limit of the array's physical aperture, the main lobe generated by the CBF algorithm is relatively rounded and wide, with obvious adhesion between the spectral peaks of the two targets and a high sidelobe level. In contrast, the Capon algorithm exhibits superior super-resolution characteristics, forming two extremely sharp and independent spectral peaks. The target orientation estimation accuracy is significantly better than that of CBF, and the noise floor is suppressed to an extremely low level. Second, in low SNR environments, as the noise power increases, the performance of the CBF algorithm deteriorates sharply, the sidelobes rise significantly, the target peaks become blurred, and spurious peaks are easily generated, making accurate angle estimation difficult. The Capon algorithm, by performing inverse operations on the signal covariance matrix, can adaptively suppress the influence of noise, maintaining good sharpness and resolution, and clearly distinguishing two adjacent targets. In summary, the Capon algorithm not only breaks through the limitations of physical aperture on angular resolution, but also demonstrates excellent robustness in weakly constrained and strongly noisy environments. It can provide more accurate steering vectors for subsequent spatial filtering, ensuring high-fidelity extraction of weak vital signs signals under complex interference.

[0043] Step 103: Using the target range gate as the center, select multiple consecutive range gates to form a region of interest. At the same time, perform static clutter suppression and phase demodulation on the signal of each range gate in the region of interest to obtain the phase sequence of each range gate.

[0044] For example, static clutter suppression uses a circle fitting algorithm. Phase demodulation uses the Differential And Cross-Multiply (DACM) algorithm.

[0045] Considering the physical thickness of the human thoracic cavity, multiple consecutive range gates centered on the target are selected as regions of interest. For the signal of each range gate within the region of interest, a hybrid model containing dynamic and static components is established. A circle fitting algorithm is used to estimate and remove the static clutter vector, correcting the center offset of the signal in the complex plane.

[0046] For example, continuous [data / text] can be selected based on the physical thickness of the human thoracic cavity and the radar resolution. Each range gate constitutes a region of interest. For the signal from each range gate within the region of interest, static clutter suppression is first performed. Static clutter, caused by environmental reflections, interferes with vital sign signals. A circular fitting algorithm is used to estimate the static clutter vector: the I / Q signals (in-phase / quadrature signals) of each range gate are modeled as circular trajectories on the complex plane, and the center of the circle is fitted using the least squares method. The fitted center is... Then, phase demodulation is performed on the denoised signal. To avoid the phase jump problem of the traditional arctangent method, phase increment is calculated using (DACM).

[0047] In one possible implementation, the circle fitting algorithm estimates the center of the circle by minimizing the sum of squared algebraic distances from the data points to the fitted circle. Its linear least squares solution is: ; ; in, The design matrix is ​​constructed based on the real and imaginary parts of the signal. The observation vector is constructed based on the real and imaginary parts of the signal; Indicates auxiliary parameters, Given the radius of the circle, estimate the center of the circle. This is the static clutter vector. The x-coordinate of the estimated center of the circle. The ordinate is the estimated center of the circle.

[0048] For example, to address the problem that traditional arctangent methods are susceptible to phase unwrapping errors caused by noise, the DAM algorithm is adopted. It directly calculates and accumulates the instantaneous phase change rate using the differential properties of the signal, achieving continuous and seamless phase demodulation.

[0049] In one possible embodiment, its discrete phase increment The calculation formula is: ; in, and They are time points The real and imaginary parts of the denoised signal.

[0050] Through accumulation This yields a continuous phase sequence for each distance gate. ;in, This represents the distance gate index.

[0051] Understandably, to verify the effectiveness of the DAM algorithm in non-stationary signal processing, embodiments of this application performed conventional arctangent demodulation and DAM algorithm processing on the same segment of radar echo data containing large-amplitude breathing motions, and the comparison results are as follows. Figure 3 As shown.

[0052] Figure 3 This is a comparison diagram of the effects of different phase extraction algorithms under large respiratory movements provided in one embodiment of this application; Figure 3 In the diagram, (a) represents the phase waveform extracted by conventional arctangent demodulation. Figure 3 In the diagram, (b) represents the phase waveform extracted based on the DAM algorithm. For example... Figure 3 As shown in (a), due to the large amplitude of human respiratory movements, the phase change exceeds the main value range, resulting in significant sawtooth-like jumps in the waveform, severely disrupting the time-domain continuity of the signal. In contrast, Figure 3 In (b), the phase waveform processed by the DAM algorithm is smooth and continuous, successfully restoring the true breathing sinusoidal shape, and high-frequency noise is effectively suppressed.

[0053] Step 104: Weighted fusion is performed based on the phase consistency measure between each phase sequence to obtain enhanced vital sign signals.

[0054] In one possible embodiment, step 104 includes: calculating the median reference sequence of all range-gate phase sequences; using the normalized cross-correlation coefficient between each phase sequence and the median reference sequence as a phase consistency measure, calculating the fusion weight based on the phase consistency measure, and performing a weighted summation of the multi-range-gate phase signals to obtain an enhanced vital sign signal.

[0055] Understandably, after obtaining the phase sequences of each range gate within all regions of interest, the quality of each phase sequence will vary due to noise and multipath effects. To enhance vital sign signals, a phase consistency metric between each phase sequence and the median reference sequence can be calculated.

[0056] In one possible embodiment, the median reference sequence of all distance-gate phase sequences is denoted as... , ;in, Indicates the total number of doors at all distances; The normalized cross-correlation coefficient between each phase sequence and the median reference sequence is used as a measure of phase consistency, and its calculation formula is as follows: No. Consistency measure of phase sequence of each distance gate with median reference sequence for: ; in, Indicates the length of the processing window. For the first A distance gate at time phase sequence, For this phase sequence in the window The mean within, Median reference sequence In the window The mean within; Assigned to the Fusion weights of phase sequences of distance gates The calculation formula is as follows: ; in, Indicates all Phase consistency metric of distance gates Perform summation; For example, if a certain channel's A value close to 1 can be assigned a fusion weight. High weight value; if a certain channel's Smaller values ​​can be assigned fusion weights. Low weight value; The weighted summation of the multi-range gate phase signals yields an enhanced vital sign signal. The calculation formula is as follows: .

[0057] Understandably, this weighted fusion method can automatically focus on high-quality signals and suppress inconsistent noise units.

[0058] Step 105: Perform spectral analysis on the enhanced vital signs signal and adaptively estimate the respiratory fundamental frequency. At the same time, use a cascaded harmonic notch filter to filter out the respiratory fundamental frequency and its harmonic components to obtain the signal after filtering out respiratory interference.

[0059] For example, enhanced vital signs signals may include respiratory and heartbeat components; since respiratory amplitude is large and harmonics may mask the heartbeat. First, spectral analysis can be performed on the enhanced vital signs signals, and then a two-stage strategy can be used to estimate the respiratory fundamental frequency.

[0060] In one possible embodiment, the respiratory fundamental frequency is adaptively estimated. A two-stage strategy (a coarse-to-fine two-stage strategy) is adopted: In the first stage (i.e., coarse estimation), the maximum value of the spectrum is searched within the preset respiratory rate range to obtain the coarsely estimated frequency. ; For example, the preset breathing rate range can be [0.1Hz, 0.5Hz].

[0061] In the second stage (i.e., fine estimation), the frequency that maximizes the cumulative energy of the fundamental frequency and its harmonics within the neighborhood of the coarse estimation frequency is searched as the finely estimated breathing fundamental frequency. Among them, accumulated energy The calculation formula is as follows: ; in, Indicating enhanced vital signs The spectrum obtained after Fast Fourier Transform Indicates the harmonic order. This represents the candidate respiratory fundamental frequencies searched by traversing the neighborhood of the coarse estimate. This represents the integral half-bandwidth of the spectrum.

[0062] Then, cascaded harmonic notch filters are designed, such as multiple sets of cascaded infinite impulse response (IIR) notch filters, each targeting the respiratory fundamental frequency. And its second and third harmonics. Each notch filter produces deep attenuation at a specific frequency to filter out breathing interference, resulting in a signal after breathing interference is removed. It mainly retains the heartbeat component.

[0063] Step 106: A noise suppression differentiator based on the weighted least squares principle is used to differentiate the signal after filtering out respiratory interference, and the displacement signal is converted into an acceleration signal to output a high-fidelity heartbeat signal.

[0064] It's understandable that the heartbeat signal is essentially a displacement signal, but the acceleration characteristics of the heartbeat are more pronounced. To highlight these characteristics, the signal after filtering out respiratory interference can be processed. Differentiation is performed. Since direct differentiation amplifies noise, a noise suppression differentiator based on the weighted least squares principle can be used. This differentiator smooths the noise using a weighted sliding window (e.g., length N=7), as illustrated in the following example.

[0065] In one possible embodiment, a noise suppression differentiator based on the weighted least squares principle performs differentiation processing on the signal after filtering out breathing interference, and its second-order differential estimation... The formula is: ; in, Indicates the filter length. Indicates the center of symmetry. Indicates the sampling time interval of the signal. This represents the weighting coefficient for the current calculation point. The signal value at the current calculation point. For the first The weighting coefficients of the symmetrical sampling points, and These represent the points relative to the current calculation point, i.e., the i-th... The first and forward The signal value at each sampling point.

[0066] Understandably, the displacement signal is converted into an acceleration signal using second-order derivatives, and the transient impact characteristics of the heartbeat are used to achieve physical-level signal enhancement. At the same time, high-frequency residual noise is smoothed through a weighted sliding window, and finally a high-fidelity heartbeat signal is output. This signal has low waveform distortion and is suitable for heart rate and rhythm analysis.

[0067] This application proposes a method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing. First, millimeter-wave radar is used to acquire echo signals containing human targets. A fast Fourier transform (FFT) is performed on the echo signals in the range dimension to obtain a range-slow-time matrix, enabling non-contact detection of distant human targets and converting the signal to a dimension capable of distinguishing targets at different distances. Then, based on the range-slow-time matrix, energy accumulation is performed within a micro-motion frequency band to determine the target range gate where the human target is located. An adaptive beamforming algorithm is then used at the target range gate to perform spatial filtering and enhancement on the multi-channel signals. This improves the robustness and accuracy of target range gate detection through energy accumulation within the frequency band, avoiding positioning failures caused by single-frame signal fluctuations. Simultaneously, adaptive beamforming achieves spatial focusing in the target direction, significantly improving the echo signal-to-noise ratio. Finally, multiple consecutive range gates are selected as the region of interest (ROI) centered on the range gate where the human target is located. Static clutter suppression and phase demodulation are then performed on the signal of each range gate within the ROI to obtain... The phase sequences of each distance gate are weighted and fused based on the phase consistency metric between the phase sequences to obtain an enhanced vital sign signal. This method utilizes the extended target characteristics of the human chest cavity to transform a single signal source into multiple signal sources, achieving spatial diversity. Furthermore, phase consistency weighted fusion yields a stable vital sign signal. Next, spectral analysis is performed on the enhanced vital sign signal, and the respiratory fundamental frequency is adaptively estimated. Simultaneously, a cascaded harmonic notch filter is used to filter out the respiratory fundamental frequency and its harmonic components. This ensures accurate estimation of the respiratory fundamental frequency, avoiding incomplete filtering or signal damage caused by estimation errors. Filtering out the respiratory fundamental frequency and its main harmonics also prevents respiratory interference with the heartbeat frequency band. Finally, a noise suppression differentiator based on the weighted least squares principle is used to differentiate the signal after filtering out respiratory interference, converting the displacement signal into an acceleration signal and outputting a high-fidelity heartbeat signal. Based on this, the proposed solution effectively addresses the technical problems of difficult heartbeat signal extraction and waveform distortion in complex environments, thereby achieving high-fidelity heartbeat waveform reconstruction.

[0068] In summary, the embodiments of this application have the following beneficial effects: Compared to traditional methods that heavily rely on close-range, direct-facing detection, the embodiments of this application effectively overcome the bottleneck of weak signal extraction in low signal-to-noise ratio (SNR) environments by deeply fusing multi-dimensional enhancement techniques across spatial, temporal, and frequency domains. The embodiments of this application first utilize Capon beamforming technology to perform directional weighted fusion of multi-channel signals from the antenna array, achieving spatial filtering enhancement in the target direction. Then, innovatively utilizing the extended target characteristics of the human chest cavity, a cascaded improvement in SNR is achieved through coherent fusion of multi-range gate signals. This multi-level signal enhancement mechanism enables the system to accurately monitor weak heartbeat signals even in weakly constrained scenarios such as long-distance and non-direct-facing detection, significantly improving the environmental adaptability and robustness of non-contact sensing.

[0069] The steps described above are for clarity only. In implementation, they can be combined into one step, or some steps can be broken down into multiple steps, as long as they involve the same logical relationship, they are all within the scope of protection of this application. Adding insignificant modifications or introducing insignificant designs to the algorithm or process, without changing the core design of the algorithm and process, are also within the scope of protection of this application.

[0070] Another embodiment of this application proposes a scene weakly constrained heartbeat signal extraction system based on millimeter-wave sensing. The details of this embodiment's scene weakly constrained heartbeat signal extraction system are described below. The following implementation details are provided for ease of understanding and are not essential for implementing this example. Figure 2 This is a schematic diagram of a scene weakly constrained heartbeat signal extraction system based on millimeter-wave sensing proposed in this embodiment, including: The signal acquisition and preprocessing module 210 is used to acquire echo signals containing human targets using millimeter-wave radar, and to perform range-dimensional fast Fourier transform on the echo signals to obtain a range-slow time matrix. The target localization and filtering enhancement module 220 is used to accumulate energy in the micro-motion frequency band based on the range-slow time matrix to determine the target range gate where the human target is located, and to use an adaptive beamforming algorithm to perform spatial filtering and enhancement on the multi-channel signal at the target range gate; The clutter suppression and phase demodulation module 230 is used to select multiple consecutive range gates to form a region of interest centered on the target range gate, and simultaneously perform static clutter suppression and phase demodulation on the signal of each range gate within the region of interest to obtain the phase sequence of each range gate; Phase consistency weighted fusion module 240 is used to perform weighted fusion based on the phase consistency measure between each phase sequence to obtain enhanced vital sign signals; The respiratory harmonic adaptive notch filter module 250 is used to perform spectral analysis on the enhanced vital signs signal and adaptively estimate the respiratory fundamental frequency, while using cascaded harmonic notch filters to filter out the respiratory fundamental frequency and its harmonic components. The signal output module 260 is used to perform differential processing on the signal after filtering out breathing interference using a noise suppression differentiator based on the weighted least squares principle, and convert the displacement signal into an acceleration signal to output a high-fidelity heartbeat signal.

[0071] It is not difficult to see that this embodiment is a system embodiment corresponding to the above method embodiments, and this embodiment can be implemented in conjunction with the above method embodiments. The relevant technical details and technical effects mentioned in the above method embodiments are still valid in this embodiment, and will not be repeated here to reduce repetition. Accordingly, the relevant technical details mentioned in this embodiment can also be applied to the above method embodiments.

[0072] It is worth mentioning that all modules and units involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units do not exist in this embodiment.

[0073] Another embodiment of this application provides an electronic device, such as Figure 3 As shown, it includes a processor 31 and a memory 32. The memory 32 stores instructions that the processor 31 can execute. When the processor 31 is configured to execute the instructions, the electronic device can realize a scene weak constraint heartbeat signal extraction method based on millimeter wave sensing as described in the above method embodiment.

[0074] The memory and processor are connected via a bus, which includes any number of interconnecting buses and bridges, connecting various circuits of one or more processors and the memory. The bus can also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single component or multiple components, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0075] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0076] Another embodiment of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, can implement a method for extracting weakly constrained heartbeat signals based on millimeter-wave sensing as described in the above method embodiments.

[0077] That is, those skilled in the art will understand that all or part of the steps in the above method embodiments can be implemented by a program instructing related hardware. The program is stored in a storage medium and includes several instructions to cause a device (such as a microcontroller, chip, etc.) or processor to execute all or part of the steps of the method described in the method embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory, random access memory, magnetic disks, or optical disks.

[0078] Those skilled in the art will understand that the above embodiments are specific implementations of this application, and in practical applications, various changes can be made in form and detail without departing from the spirit and scope of this application. For those skilled in the art, several improvements and modifications can be made without departing from the principles of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.

Claims

1. A method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing, characterized in that, The method includes: Millimeter-wave radar is used to collect echo signals containing human targets, and the echo signals are subjected to range-dimensional fast Fourier transform to obtain the range-slow time matrix. Based on the range-slow time matrix, energy accumulation is performed within the micro-motion frequency band to determine the target range gate where the human target is located, and an adaptive beamforming algorithm is used at the target range gate to perform spatial filtering and enhancement on the multi-channel signal; Centered on the target range gate, multiple consecutive range gates are selected to form a region of interest. At the same time, static clutter suppression and phase demodulation are performed on the signal of each range gate within the region of interest to obtain the phase sequence of each range gate. Weighted fusion based on phase consistency measurement between phase sequences yields enhanced vital sign signals. The enhanced vital signs signal was subjected to spectral analysis, and the respiratory fundamental frequency was adaptively estimated. At the same time, a cascaded harmonic notch filter was used to filter out the respiratory fundamental frequency and its harmonic components, resulting in a signal after filtering out respiratory interference. A noise suppression differentiator based on the weighted least squares principle is used to differentiate the signal after filtering out respiratory interference, and the displacement signal is converted into an acceleration signal to output a high-fidelity heartbeat signal.

2. The method according to claim 1, characterized in that, The adaptive beamforming algorithm is the Capon beamforming algorithm; the optimal weight vector of the Capon beamforming algorithm. The calculation formula is: ; in, The covariance matrix of the received signal. For scanning angle The corresponding guide vector, express The conjugate transpose of .

3. The method according to claim 1, characterized in that, Static clutter suppression employs a circle fitting algorithm, which estimates the center of the circle by minimizing the sum of squared algebraic distances from data points to the fitted circle. Its linear least squares solution is: ； ； in, The design matrix is ​​constructed based on the real and imaginary parts of the signal. The observation vector is constructed based on the real and imaginary parts of the signal; Indicates auxiliary parameters, Given the radius of the circle, estimate the center of the circle. This is the static clutter vector.

4. The method according to claim 1, characterized in that, Phase demodulation employs a differential and cross-multiplication algorithm, with its discrete phase increment... The calculation formula is: ； in, and They are time points The real and imaginary parts of the denoised signal.

5. The method according to claim 1, characterized in that, The weighted fusion based on the consistency metric between each phase sequence yields enhanced vital sign signals, including: Calculate the median reference sequence for all range-gated phase sequences; The normalized cross-correlation coefficients of each phase sequence and the median reference sequence are used as a phase consistency measure. The fusion weights are calculated based on the phase consistency measure, and the multi-range gate phase signals are weighted and summed to obtain the enhanced vital signs signal.

6. The method according to claim 5, characterized in that, Let the median reference sequence of all distance gate phase sequences be . , ;in, Indicates the total number of doors at all distances; The normalized cross-correlation coefficient between each phase sequence and the median reference sequence is used as a measure of phase consistency, and its calculation formula is as follows: No. Consistency measure of distance gates for: ； in, Indicates the length of the processing window. For the first A distance gate at time phase sequence, For this phase sequence in the window The mean within, Median reference sequence In the window The mean within; Assigned to the Fusion weights of phase sequences of distance gates The calculation formula is as follows: ； in, Indicates all Phase consistency metric of distance gates Perform summation; The weighted summation of the multi-range gate phase signals yields an enhanced vital sign signal. The calculation formula is as follows: 。 7. The method according to claim 6, characterized in that, Adaptive estimation of respiratory fundamental frequency A two-stage strategy is adopted: In the first stage, the maximum value of the spectrum within the preset respiratory rate range is searched to obtain a coarse estimate of the frequency. ; In the second stage, the frequency that maximizes the cumulative energy of the fundamental frequency and its harmonics within the neighborhood of the coarsely estimated frequency is searched as the finely estimated respiratory fundamental frequency. Among them, accumulated energy The calculation formula is as follows: ； in, Indicating enhanced vital signs The spectrum obtained after Fast Fourier Transform Indicates the harmonic order. This represents the candidate respiratory fundamental frequencies searched by traversing the neighborhood of the coarse estimate. This represents the integral half-bandwidth of the spectrum.

8. The method according to claim 7, characterized in that, A noise suppression differentiator based on the weighted least squares principle performs differentiation on the signal after filtering out breathing interference, and its second-order differential estimation... The formula is: ； in, Indicates the filter length. Indicates the center of symmetry. Indicates the sampling time interval of the signal. This represents the weighting coefficient for the current calculation point. The signal value at the current calculation point. For the first The weighting coefficients of the symmetrical sampling points, and These represent the points relative to the current calculation point, i.e., the i-th... The first and forward The signal value at each sampling point.

9. An electronic device, characterized in that, include: The processor and memory, wherein the memory stores instructions that the processor can execute, and the processor is configured to, when executing the instructions, enable the electronic device to implement a method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it can implement a method for extracting weakly constrained heartbeat signals in a scene based on millimeter-wave sensing as described in any one of claims 1 to 8.

Images