Millimeter wave radar vital sign signal processing system and method

Abstract

The invention relates to the technical field of millimeter wave radar vital sign detection, and discloses a millimeter wave radar vital sign signal processing system and method. According to the method, distance dimension processing and phase extraction are carried out on millimeter wave radar echo signals, heartbeat signals are enhanced through phase difference, and the time domain sparse characteristic of random motion interference is highlighted; then, a sparse optimization model combining a time domain and a frequency domain is constructed, differential sparse constraints are applied to a breathing frequency band, a heartbeat frequency band and a heartbeat harmonic frequency band in the frequency domain, and two-norm morphological constraints are applied to sparse peaks of random motion interference in the time domain; and finally, the model is solved to realize separation of interference and vital sign signals. Complex hardware configuration is not needed, weak heartbeat signal error suppression can be effectively avoided, the stability and accuracy of respiratory rate and heart rate estimation in a random human body motion scene are remarkably improved, and the method is suitable for a non-contact vital sign monitoring scene.

Description

Technical Field

[0001] This invention relates to a millimeter-wave radar vital sign signal processing system and method, which is particularly suitable for scenarios that suppress random human motion interference and stably extract respiratory and heartbeat signals, and belongs to the field of radar signal processing and non-contact vital sign detection technology. Background Technology

[0002] Respiratory rate and heart rate are core parameters reflecting the physiological state of the human body, and they have irreplaceable application value in fields such as medical monitoring, home health management, sleep monitoring, and vehicle occupant status perception. Traditional vital sign monitoring relies on contact devices such as ECG electrodes and breathing belts, which have inherent drawbacks such as discomfort when wearing them, poor compliance, and inability to conduct long-term continuous monitoring, making it difficult to meet the diversified needs of modern health monitoring.

[0003] With the development of radar sensing technology, non-contact vital sign detection methods based on millimeter-wave radar have emerged. Millimeter-wave radar, with its high frequency, short wavelength, and high phase sensitivity, can accurately capture the minute displacements of the human chest cavity caused by breathing and heartbeat, achieving long-distance, non-contact vital sign detection, and has gradually become a research hotspot in this field. However, in practical applications, the object being measured is unlikely to remain absolutely still. Random human movements such as swaying, posture adjustments, and limb movements introduce large-amplitude, unevenly distributed phase disturbances into the radar echo. These disturbances often have energy far exceeding that of weak vital sign signals such as heartbeats, easily causing target range cell drift, phase abrupt changes, and spectral aliasing, severely damaging the integrity of vital sign signals, leading to distorted detection results and a significant decrease in reliability.

[0004] To address the problem of random human motion interference, existing technologies have proposed various solutions, but all have significant limitations: 1. Multi-sensor fusion solutions (such as combining radar and cameras): compensate for motion interference through the collaboration of multiple devices, but the system is complex and the hardware cost is high. In addition, the use of cameras poses a risk of privacy leakage, which limits its application in private scenarios such as home and medical care. 2. Multi-antenna structure scheme: It uses spatial signal differences to eliminate interference, but it has strict requirements for hardware configuration and requires customized antenna arrays, which is not conducive to low-cost large-scale deployment; 3. Time-frequency analysis methods (such as empirical mode decomposition and wavelet transform): These methods separate interference from vital signs through signal decomposition. However, when the amplitude of random human movement is large, the interference signal is easily misdecomposed into effective modes, masking or destroying weak heartbeat signals, resulting in poor separation performance. 4. Traditional sparse optimization methods: These methods utilize the sparsity of motion disturbances for suppression. However, by employing a unified global constraint strategy, which prioritizes the retention of high-energy respiratory or motion components, they are prone to misjudging weak heartbeat signals as noise and removing them, leading to unstable and severely distorted heart rate estimation.

[0005] In summary, existing technologies for suppressing random human motion interference either rely on complex hardware, leading to high costs and poor applicability, or suffer from ineffective protection of weak heartbeat signals due to unreasonable signal processing strategies, making it difficult to achieve a balance between "hardware simplicity" and "detection accuracy." Therefore, how to accurately separate random human motion interference from vital sign signals without significantly increasing hardware complexity, and avoid false suppression of weak heartbeat signals, has become a core technical problem that urgently needs to be solved in the field of millimeter-wave radar non-contact vital sign detection. Summary of the Invention

[0006] To overcome the problems of large phase perturbation amplitude, masking of weak heartbeat signals, and unstable detection results in existing millimeter-wave radar vital sign detection methods when random human movement is present, this invention provides a millimeter-wave radar vital sign signal processing system and method. The aim is to improve the stability and reliability of vital sign parameter extraction, such as respiratory rate and heart rate, without significantly increasing hardware complexity. This system achieves efficient separation of random human movement interference from respiratory and heartbeat signals, thereby improving the stability and accuracy of vital sign parameter estimation.

[0007] A millimeter-wave radar vital sign signal processing system and method includes the following steps: The signal acquisition module is used to acquire millimeter-wave radar echo signals. The signal acquisition module adopts frequency-modulated continuous wave millimeter-wave radar, periodically transmits linear frequency-modulated continuous wave signals, receives echo signals reflected by the human body, and converts them into digital baseband echo signals. The phase processing module is used to perform range dimension processing and phase extraction on the echo signal; the module performs fast time dimension fast Fourier transform, target range cell selection based on energy criterion, and phase demodulation based on differential and cross multiplication; The differential enhancement module is used to perform first-order differential processing and low-pass filtering on the phase signal, enhance the high-frequency components of the heartbeat signal, and highlight the time-domain sparsity characteristics of random human motion interference. The sparse optimization module is used to separate random human motion interference and vital sign signals based on a sparse optimization model in the joint time and frequency domains. The module constructs an optimization model that combines piecewise sparse constraints and morphological constraints, and solves it through a fast iterative threshold algorithm. The vital signs output module is used to output vital signs signals after removing random human motion interference, and to extract respiratory rate and heart rate parameters through spectrum analysis.

[0008] A processing method for a millimeter-wave radar vital sign signal processing system includes the following steps: Step 1: Perform range dimension processing on the millimeter-wave radar echo signal to obtain the complex signal corresponding to the target range cell; Step 2: Perform phase extraction on the complex signal to obtain a phase signal that varies with time; Step 3: Perform differential processing on the phase signal to enhance the high-frequency components of the heartbeat signal and highlight the sparse characteristics of random human motion interference in the time domain. Step 4: Construct a sparse optimization model in the joint time and frequency domains based on the differentially processed signal. Apply differentiated sparse constraints to different physiological signal frequency bands in the frequency domain, and apply morphological constraints to the sparse peaks corresponding to random human motion interference in the time domain. Step 5: By solving the sparse optimization model, the random human motion interference and vital sign signals are separated to obtain the interference-free vital sign signals.

[0009] This invention provides a processing method for a millimeter-wave radar vital sign signal processing system. First, the millimeter-wave radar echo signal undergoes range dimension processing, selecting the range cell corresponding to the target and extracting its complex signal. Based on this, a time-varying phase signal is obtained through phase demodulation, and differential processing is performed on the phase signal to enhance the high-frequency components corresponding to the heartbeat signal, while simultaneously making random human motion interference exhibit more pronounced sparsity characteristics in the time domain. Subsequently, a joint time-domain and frequency-domain sparse optimization model is constructed based on the differentially processed signal. In the frequency domain, the signal is segmented and modeled according to the breathing frequency band, heartbeat frequency band, and heartbeat harmonic frequency band, and different sparsity constraint strengths are set for different frequency bands to avoid the weak heartbeat signal being mistakenly suppressed during the global sparsity optimization process. Simultaneously, morphological constraints are introduced in the time domain for the sparse peak positions corresponding to random human motion interference to maintain the consistency of interference characteristics and improve signal separation accuracy. By solving the sparse optimization model, effective separation of random human motion interference and vital sign signals is achieved, obtaining the interference-free vital sign signal.

[0010] The range dimension processing in step one includes performing a fast time-dimensional fast Fourier transform on the millimeter-wave radar echo signal and selecting target range cells based on energy criteria. Specifically, a fast Fourier transform is performed on the fast time-dimensional sampling sequence of each frame of echo signal to convert the time-domain signal into a range-domain spectrum. The amplitude spectrum of each range cell at all slow-time sampling points is accumulated to obtain the energy value of each range cell, and the range cell with the highest energy is selected as the target range cell.

[0011] The phase extraction in step two adopts a phase demodulation method based on differential and cross multiplication to avoid the phase discontinuity problem introduced by arctangent demodulation. Let the complex signal of the target range cell in the slow time dimension be... ,in , The first The calculation process for phase demodulation of the in-phase and quadrature components at each slow-time sampling point is as follows: Phase information that changes over time is obtained by accumulating phase changes.

[0012] In step three, the differential processing involves performing a first-order differential operation on the phase signal, followed by low-pass filtering to suppress high-frequency noise related to non-vital signs. The calculation process for the first-order differential operation is as follows: ,in This is the signal after differential processing.

[0013] In step four, the frequency domain sparsity constraint divides the signal into the respiratory frequency band (0.1Hz–0.6Hz), the heartbeat frequency band (0.8Hz–3Hz), the heartbeat harmonic frequency band (3Hz–6Hz), and other frequency bands according to the frequency range, and sets different sparsity penalty coefficients for different frequency bands; among them, the sparsity penalty coefficients of the heartbeat frequency band and the heartbeat harmonic frequency band are smaller than those of the respiratory frequency band and other frequency bands.

[0014] In step four, the temporal morphological constraint is applied to the temporal peak position corresponding to random human motion interference. An error constraint based on the second norm is introduced to maintain the consistency of the amplitude and shape of the interference peak. Specifically, the position of the random motion peak in the differential signal is identified by threshold detection, and the second norm constraint is applied to the interference component at the position so that the optimization result is consistent with the amplitude and shape of the original observed signal at the peak position.

[0015] A computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the millimeter-wave radar vital sign signal processing method described in any of the preceding claims.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention combines phase differential processing and piecewise sparsity optimization strategies to distinguish and model random human motion interference and vital sign signals in both the time and frequency domains. This fully utilizes the sparsity characteristics of random human motion interference and the spectral characteristics of vital sign signals, thus maintaining the separability of respiratory and heartbeat signals even in the presence of significant random human motion. This invention does not rely on multi-sensor fusion or complex antenna structures and is suitable for non-contact vital sign monitoring applications.

[0017] This invention addresses the problems of large phase disturbance amplitude, strong temporal sparsity, and easy suppression of weak signals such as heartbeats caused by random human movement during non-contact vital sign detection. It proposes an interference suppression method based on a combination of differential processing and two-layer sparse optimization. First, the method performs range dimension processing and phase extraction on the millimeter-wave radar echo signal, enhancing the heartbeat signal and highlighting the temporal sparsity of random motion interference through phase differential operation. Then, it segments and models the signal in the frequency domain, setting different sparsity constraint strengths for the respiratory frequency band, heartbeat frequency band, and their harmonic frequency bands to avoid false suppression of weak heartbeat signals during global sparsity optimization. Simultaneously, it introduces a L2 constraint for the peak position of random motion in the time domain to maintain the consistency of interference patterns and improve separation accuracy. Through combined time-domain and frequency-domain two-layer optimization, effective separation of random human movement interference and vital sign signals is achieved. This invention improves the stability and accuracy of respiratory rate and heart rate estimation even in the presence of significant random human movement, making it suitable for non-contact vital sign monitoring scenarios. The combination of differential processing and sparsity enhancement lays the foundation for efficient separation: by performing first-order differential operation on the phase signal, the high-frequency components of the heartbeat signal are amplified, increasing their energy proportion in the signal; on the other hand, random human motion interference is transformed into isolated spikes in the time domain, significantly enhancing its sparsity characteristics. This provides a clear structural feature basis for subsequent interference suppression based on sparse modeling, solving the problem that weak heartbeat signals are easily masked by strong interference.

[0018] Frequency domain segmentation and differentiated sparsity constraints avoid false suppression of weak heartbeat signals: Based on the frequency differences and energy distribution characteristics of respiration, heartbeat, and heartbeat harmonics, dedicated frequency bands are divided and differentiated sparsity penalty coefficients are set. More lenient constraints are applied to the weaker heartbeat and harmonic frequency bands, and their signal components are retained first. This completely solves the problem of false suppression of heartbeat signals caused by the "emphasis on strong and neglect of weak" in traditional global sparsity optimization, and significantly improves the accuracy of heart rate estimation.

[0019] Temporal morphological constraints improve separation accuracy: A second-norm constraint is introduced to address the spike characteristics of random motion interference. This constraint suppresses interference while maintaining the consistency of its amplitude and shape, avoiding spike distortion or residual interference caused by simple first-norm constraints. This improves the separation accuracy between interference and vital signs signals, ensuring the integrity of vital signs signals.

[0020] The hardware structure is simple, low-cost and easy to deploy: it can be implemented based solely on single-frequency continuous wave millimeter-wave radar without the need for additional sensors, cameras or complex antenna arrays, which reduces hardware costs and system complexity. It is suitable for large-scale applications in various scenarios such as medical monitoring, home health, and vehicle perception, while avoiding the privacy leakage risks brought about by multi-sensor fusion.

[0021] Strong anti-interference capability and stable estimation results: Through time-domain-frequency domain dual-layer optimization modeling, the sparsity characteristics of interference and the spectral characteristics of vital signs are fully utilized. Even in scenarios with large random human movement amplitude, breathing and heartbeat signals can still be stably separated. The fluctuation amplitude of breathing frequency and heart rate estimation is significantly reduced, which significantly improves the reliability of non-contact vital sign monitoring. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the overall structure of a millimeter-wave radar vital sign signal processing system and method according to the present invention; Figure 2 This is a schematic diagram of the data structure of the millimeter-wave radar echo signal in the range dimension (fast time dimension) and the slow time dimension of the present invention; Figure 3 This is a schematic diagram comparing the time-domain waveforms of the phase signal before and after differential processing according to the present invention (the top is the original phase signal, and the bottom is the signal after differential processing; it can be seen that the high-frequency components of the heartbeat are enhanced, and the motion interference exhibits peak-sparse characteristics). Figure 4 This is a schematic diagram of the characteristic distribution of random human motion interference and vital sign signals in the frequency domain of the present invention (labeled with the frequency range and energy difference of the respiratory frequency band, heartbeat frequency band, and heartbeat harmonic frequency band). Figure 5 This is a schematic diagram of the random human motion interference suppression process based on joint time-domain and frequency-domain sparse optimization of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Reference Figure 1 A millimeter-wave radar vital sign signal processing system and method, comprising the following steps: The signal acquisition module is used to acquire millimeter-wave radar echo signals. The signal acquisition module adopts frequency-modulated continuous wave millimeter-wave radar, periodically transmits linear frequency-modulated continuous wave signals, receives echo signals reflected by the human body, and converts them into digital baseband echo signals. The phase processing module is used to perform range dimension processing and phase extraction on the echo signal; the module performs fast time dimension fast Fourier transform, target range cell selection based on energy criterion, and phase demodulation based on differential and cross multiplication; The differential enhancement module is used to perform first-order differential processing and low-pass filtering on the phase signal, enhance the high-frequency components of the heartbeat signal, and highlight the time-domain sparsity characteristics of random human motion interference. The sparse optimization module is used to separate random human motion interference and vital sign signals based on a sparse optimization model in the joint time and frequency domains. The module constructs an optimization model that combines piecewise sparse constraints and morphological constraints, and solves it through a fast iterative threshold algorithm. The vital signs output module is used to output vital signs signals after removing random human motion interference, and to extract respiratory rate and heart rate parameters through spectrum analysis.

[0026] A processing method for a millimeter-wave radar vital sign signal processing system includes the following steps: Step 1: Perform range dimension processing on the millimeter-wave radar echo signal to obtain the complex signal corresponding to the target range cell; Step 2: Perform phase extraction on the complex signal to obtain a phase signal that varies with time; Step 3: Perform differential processing on the phase signal to enhance the high-frequency components of the heartbeat signal and highlight the sparse characteristics of random human motion interference in the time domain. Step 4: Construct a sparse optimization model in the joint time and frequency domains based on the differentially processed signal. Apply differentiated sparse constraints to different physiological signal frequency bands in the frequency domain, and apply morphological constraints to the sparse peaks corresponding to random human motion interference in the time domain. Step 5: By solving the sparse optimization model, the random human motion interference and vital sign signals are separated to obtain the interference-free vital sign signals.

[0027] In this embodiment, a frequency-modulated continuous wave (FM-CW) millimeter-wave radar is used as the signal acquisition device to obtain radar echo signals related to human vital signs. The millimeter-wave radar periodically transmits FM-CW signals to the target human body according to preset operating parameters. The transmitted signals are reflected upon encountering the surface of the human chest cavity during propagation, and the reflected signals contain minute displacement modulation information caused by respiration and heartbeat. The radar receiver performs down-conversion processing on the reflected echoes and obtains digitized baseband echo signals after analog-to-digital conversion.

[0028] During the digital processing stage, the radar echo signal is organized in two dimensions according to the radar's operating time sequence. One dimension corresponds to the sampling sequence within a single linear frequency modulated (LFM) signal, used to characterize the target's range information, and is usually referred to as the fast time dimension. The other dimension corresponds to the transmission and reception process of multiple consecutive LFM signals, used to characterize the signal's change over time, and is usually referred to as the slow time dimension. Thus, the radar echo signal is constructed as a two-dimensional complex signal matrix composed of multiple slow-time sampling points and multiple fast-time sampling points. Each row of the matrix corresponds to the echo data of a single LFM signal, and each column corresponds to the signal change of the same range unit at different times.

[0029] The aforementioned two-dimensional data structure allows for the extraction of distance information in the fast time dimension and the analysis of phase changes caused by human vital signs in the slow time dimension during subsequent processing. This provides a data foundation for subsequent signal processing steps such as distance unit selection, phase demodulation, and suppression of random human motion interference.

[0030] First, range dimension processing is performed on each frame of radar echo signal in the fast time dimension to extract echo features corresponding to different range cells. In this embodiment, the range dimension processing is achieved by performing a Fast Fourier Transform on the fast time dimension signal, thereby converting the time-domain sampled signal into range-domain spectral information. Let the... The radar echo signal acquired at each slow-time sampling point is represented as follows: Where n represents the slow-time sampling index, This indicates a fast-time sampling index, and The value range is 0 to Then, by performing a Fast Fourier Transform on the echo signal in the fast time dimension, the corresponding range domain transform result can be obtained. Its mathematical expression is: in This represents the distance cell index, corresponding to different target distance locations.

[0031] After the aforementioned distance dimension transformation, the amplitude changes of each distance cell in the slow-time dimension reflect the reflection intensity of the target at different distance positions. To stably select the distance cell containing the target human body, this embodiment accumulates the amplitude spectrum of each distance cell across all slow-time sampling points to obtain the energy value of the corresponding distance cell. By comparing the accumulated energy of each distance cell, the distance cell corresponding to the target human body is determined based on the maximum energy criterion, thereby avoiding the influence of instantaneous noise or short-term disturbances on the distance cell selection result.

[0032] After determining the target distance unit, the complex signal sequence of the distance unit in the slow time dimension is extracted from the two-dimensional complex signal matrix. The complex signal sequence contains the small displacement modulation information of the target human body caused by breathing and heartbeat during the observation time, and serves as the input signal for subsequent signal processing steps such as phase demodulation, differential processing, and random human motion interference suppression.

[0033] Subsequently, the complex signal at the extracted target distance unit is subjected to phase demodulation to obtain phase information that changes over time. In this embodiment, to improve the stability of phase demodulation under conditions of large displacement disturbances, a phase demodulation method based on differential and cross-multiplication is adopted to estimate the phase change of the complex signal, thereby avoiding the problem of phase jumps and discontinuities that are prone to occur when the phase crosses quadrants in traditional phase demodulation methods based on the arctangent function.

[0034] Let the complex signal of the target distance unit in the slow time dimension be represented as: ,in and They represent the first The in-phase and quadrature components at each slow-time sampling point. The phase demodulation method based on differential and cross-multiplication accumulates the phase change signal by performing differential and cross-multiplication operations on the in-phase and quadrature components of adjacent sampling points. The calculation process can be expressed as: Through the above calculations, phase change information can be obtained without explicitly calculating the arctangent function, thus effectively avoiding demodulation errors caused by phase wrapping or abrupt phase changes. This phase demodulation method is highly adaptable to large displacement disturbances caused by human body swaying or posture adjustments, and is suitable for practical applications of non-contact vital sign detection.

[0035] After obtaining the phase change signal, to further enhance the heartbeat signal and highlight the characteristics of random human motion interference, a first-order difference processing is performed on the phase change signal. This first-order difference processing characterizes the rate of phase change over time, and its calculation process can be expressed as follows: This differential operation amplifies the rapidly changing components in the phase signal in the time domain, resulting in a relative enhancement of the higher-frequency heartbeat signal in the differential result, while the lower-frequency respiratory signal is suppressed to some extent.

[0036] From a frequency domain perspective, the first-order differential operation is equivalent to multiplying the spectrum of the original phase signal by a weighting factor proportional to the frequency. Therefore, high-frequency components in the differentially processed signal have a greater amplitude weight. Consequently, the energy proportion of the heartbeat signal in the differential signal is relatively increased, which is beneficial for preserving weak heartbeat signals during subsequent sparse optimization. Simultaneously, phase abrupt changes caused by random human motion manifest as isolated spikes with large amplitudes in the differential signal. These spikes exhibit a clear discontinuous distribution characteristic on the time axis, possessing strong temporal sparsity, providing significant structural features for subsequent interference suppression based on sparse modeling.

[0037] After phase differential processing, a joint time-domain and frequency-domain sparse optimization model is constructed to suppress random human motion interference and separate vital signs signals from the differential signal. This sparse optimization model comprehensively utilizes the sparsity characteristics of random human motion interference in the time domain and the structured distribution characteristics of vital signs signals in the frequency domain to achieve joint modeling and separation of different signal components.

[0038] Let the observed signal after first-order difference processing be represented as ,in The observed signal is a one-dimensional sequence arranged along the slow time dimension. It is modeled as a superposition of time-domain sparse interference components, frequency-domain vital signs components, and residual noise components, and its mathematical expression can be represented as: in, This represents the time-domain sparse disturbance component caused by random human motion. This component exhibits a peak structure with large amplitude but discontinuous distribution on the time axis. It represents the sparse coefficient vector of vital signs signals in the frequency domain, corresponding to periodic physiological activities such as breathing and heartbeat; This is the normalized Fourier transform matrix, used to map the frequency domain coefficients to the time domain; This represents the residual noise term other than the two types of components mentioned above, used to characterize system noise and unmodeled errors.

[0039] Considering the differences in frequency range and energy distribution between respiratory and heartbeat signals, to avoid the false suppression of the weaker heartbeat signal under uniform sparsity constraints, the sparsity coefficients of vital signs are adjusted in the frequency domain. Segmented modeling is performed based on frequency range. Specifically, [the model will be] segmented by frequency range. The coefficients are divided into a subset corresponding to the respiratory frequency band, a subset corresponding to the heartbeat frequency band, a subset corresponding to the heartbeat harmonic frequency band, and a subset corresponding to the remaining frequency bands. Different sparsity penalty coefficients are set for different frequency bands to achieve differentiated constraints on different vital sign components.

[0040] Based on the above modeling method, a sparse optimization objective function jointly applied in the time and frequency domains is constructed, which can be expressed as: The first term is the data consistency term, used to constrain the error between the model-reconstructed signal and the observed signal; the second term is the time-domain sparsity constraint term, which is determined by... The first norm constraint ensures that random human motion disturbances remain sparsely distributed along the time axis; the third term is a piecewise sparse constraint term in the frequency domain, where... This represents the subset of frequency domain coefficients corresponding to the i-th frequency band. This is the sparse penalty parameter corresponding to this frequency band, used to control the degree of preservation of vital sign signals in different frequency bands.

[0041] Furthermore, considering that random human motion interference typically manifests as isolated spikes with large amplitudes in differential signals, a local L2 constraint mechanism is introduced on top of the temporal sparsity constraint. Specifically, after detecting the spike positions, L2 error constraints are applied to the temporal interference components at these positions, ensuring that the optimized results maintain the same amplitude and morphological characteristics as the original observed signal at the corresponding positions. This avoids excessive compression of spike amplitudes or morphological distortion caused by simple L1 constraints. Through these improvements, random human motion interference is effectively separated while its morphological information is reasonably preserved, thereby improving the stability and accuracy of the overall interference suppression and vital sign separation process.

[0042] The aforementioned sparse optimization problem in both the time and frequency domains is solved using a fast iterative thresholding algorithm. In this embodiment, the fast iterative thresholding algorithm is implemented based on a proximal gradient optimization framework. By dividing the objective function into differentiable data consistency terms and non-differentiable sparse constraint terms, gradient updates and threshold shrinkage operations are alternately performed in each iteration, thereby achieving joint estimation of time-domain interference components and frequency-domain vital sign components.

[0043] Let the first The temporal interference component and the frequency domain vital signs component at the next iteration are respectively and First, construct the residual signal. in The observed signal after differentiation. Let be the normalized Fourier transform matrix. Based on the residual signal, gradient descent is performed on the time-domain and frequency-domain components respectively, and the update form can be expressed as: in The step size parameter is related to the Lipschitz constant of the objective function. This represents the conjugate transpose of the Fourier transform matrix.

[0044] After gradient update, a threshold shrinkage operation is applied to the update result to satisfy the sparsity constraint. For the temporal disturbance component, a threshold shrinkage operation is applied at non-peak locations. Applying the soft thresholding operator corresponding to the first norm can be expressed as follows: in This represents the soft threshold operator. For the time-domain sparse constraint parameters, a L2 norm error constraint is introduced during the update process for the identified random human motion peak positions. This ensures that the update results at the corresponding positions maintain the same amplitude and morphological characteristics as the original observed signal while minimizing the objective function.

[0045] For frequency domain vital sign components, in the frequency domain, according to the pre-defined frequency range, ... Apply segmented threshold update rules. Specifically, for the respiratory frequency band, heart rate frequency band, heart rate harmonic frequency band, and other frequency bands, different threshold parameters are used for soft threshold contraction, and the update form can be expressed as: in This represents the subset of frequency domain coefficients corresponding to the i-th frequency band. This is the sparse penalty parameter corresponding to this frequency band.

[0046] To improve the algorithm's convergence speed, this embodiment introduces an acceleration mechanism based on the results of two iterations. Let the acceleration parameter be... Then its update method can be expressed as: An auxiliary variable is constructed using a linear combination of the current iteration result and the previous iteration result for the next gradient calculation. This method reduces the overall number of iterations while maintaining algorithm stability. When the change in the result between two adjacent iterations is less than a preset threshold, or when the maximum number of iterations is reached, the algorithm is deemed to have met the convergence condition and the iteration terminates.

[0047] After solving the fast iterative thresholding algorithm, the sparse coefficients of vital signs in the frequency domain are obtained. By performing an inverse Fourier transform on these frequency domain sparse coefficients, the time-domain signal of vital signs after removing random human motion interference can be reconstructed. This reconstructed signal exhibits continuous periodic variation characteristics in the time domain, while the discontinuous interference components caused by random human motion are effectively suppressed.

[0048] In this embodiment, the time-domain signal of vital signs is further subjected to spectral analysis. The peak frequency of the dominant frequency is extracted within a preset respiratory frequency range and heart rate range, thereby obtaining the corresponding respiratory frequency and heart rate parameters. Through the above processing procedure, even in the presence of random human motion interference, stable separation and parameter extraction of vital sign signals can be achieved, making it suitable for non-contact vital sign monitoring applications based on millimeter-wave radar.

[0049] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and these variations still fall within the protection scope of the present invention.

Claims

1. A millimeter-wave radar vital sign signal processing system, characterized in that: include: The signal acquisition module is used to acquire millimeter-wave radar echo signals; The signal acquisition module uses a frequency-modulated continuous wave millimeter-wave radar, which periodically transmits linear frequency-modulated continuous wave signals, receives echo signals reflected by the human body, and converts them into digital baseband echo signals. The phase processing module is used to perform range dimension processing and phase extraction on the echo signal; the module performs fast time dimension fast Fourier transform, target range cell selection based on energy criterion, and phase demodulation based on differential and cross multiplication; The differential enhancement module is used to perform first-order differential processing and low-pass filtering on the phase signal, enhance the high-frequency components of the heartbeat signal, and highlight the time-domain sparsity characteristics of random human motion interference. The sparse optimization module is used to separate random human motion interference and vital sign signals based on a sparse optimization model in the joint time and frequency domains. The module constructs an optimization model that combines piecewise sparse constraints and morphological constraints, and solves it through a fast iterative threshold algorithm. The vital signs output module is used to output vital signs signals after removing random human motion interference, and to extract respiratory rate and heart rate parameters through spectrum analysis.

2. A processing method for a millimeter-wave radar vital sign signal processing system according to claim 1, characterized in that: Includes the following steps: Step 1: Perform range dimension processing on the millimeter-wave radar echo signal to obtain the complex signal corresponding to the target range cell; Step 2: Perform phase extraction on the complex signal to obtain a phase signal that varies with time; Step 3: Perform differential processing on the phase signal to enhance the high-frequency components of the heartbeat signal and highlight the sparse characteristics of random human motion interference in the time domain. Step 4: Construct a sparse optimization model in the joint time and frequency domains based on the differentially processed signal. Apply differentiated sparse constraints to different physiological signal frequency bands in the frequency domain, and apply morphological constraints to the sparse peaks corresponding to random human motion interference in the time domain. Step 5: By solving the sparse optimization model, the random human motion interference and vital sign signals are separated to obtain the interference-free vital sign signals.

3. The millimeter-wave radar vital sign signal processing method according to claim 1, characterized in that: The range dimension processing in step one includes performing a fast time-dimensional fast Fourier transform on the millimeter-wave radar echo signal and selecting target range cells based on energy criteria. Specifically, a fast Fourier transform is performed on the fast time-dimensional sampling sequence of each frame of echo signal to convert the time-domain signal into a range-domain spectrum. The amplitude spectrum of each range cell at all slow-time sampling points is accumulated to obtain the energy value of each range cell, and the range cell with the highest energy is selected as the target range cell.

4. The millimeter-wave radar vital sign signal processing method according to claim 1, characterized in that: The phase extraction in step two adopts a phase demodulation method based on differential and cross multiplication to avoid the phase discontinuity problem introduced by arctangent demodulation. Let the complex signal of the target range cell in the slow time dimension be... ,in , The first The calculation process for phase demodulation of the in-phase and quadrature components at each slow-time sampling point is as follows: Phase information that changes over time is obtained by accumulating phase changes.

5. The millimeter-wave radar vital sign signal processing method according to claim 1, characterized in that: In step three, the differential processing involves performing a first-order differential operation on the phase signal, followed by low-pass filtering to suppress high-frequency noise related to non-vital signs. The calculation process for the first-order differential operation is as follows: ,in This is the signal after differential processing.

6. The millimeter-wave radar vital sign signal processing method according to claim 1, characterized in that: In step four, the frequency domain sparsity constraint divides the signal into the respiratory frequency band (0.1Hz–0.6Hz), the heartbeat frequency band (0.8Hz–3Hz), the heartbeat harmonic frequency band (3Hz–6Hz), and other frequency bands according to the frequency range, and sets different sparsity penalty coefficients for different frequency bands; among them, the sparsity penalty coefficients of the heartbeat frequency band and the heartbeat harmonic frequency band are smaller than those of the respiratory frequency band and other frequency bands.

7. The millimeter-wave radar vital sign signal processing method according to claim 1, characterized in that: In step four, the temporal morphological constraint is applied to the temporal peak position corresponding to random human motion interference. An error constraint based on the second norm is introduced to maintain the consistency of the amplitude and shape of the interference peak. Specifically, the position of the random motion peak in the differential signal is identified by threshold detection, and the second norm constraint is applied to the interference component at the position so that the optimization result is consistent with the amplitude and shape of the original observed signal at the peak position.

8. A computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the millimeter-wave radar vital sign signal processing method according to any one of claims 2 to 7.

Images