FMCW radar vital sign detection method based on combination of SVMD and SSR

Abstract

The invention discloses an FMCW radar vital sign detection method based on combination of SVMD and SSR. The method comprises the following steps: firstly, acquiring an FMCW radar echo signal, and preprocessing the FMCW radar echo signal to obtain a human chest wall micro-motion phase signal; then, an improved SVMD algorithm is adopted to decompose the human body chest wall micro-motion phase signal, the optimal balance parameter upper limit value is determined through a joint evaluation criterion, and a modal component is obtained; and finally, carrying out harmonic detection and signal reconstruction to obtain a pure heartbeat signal. And constructing an SSR model, performing sparse spectrum reconstruction based on an EB-ZA-SLMS algorithm, solving a heartbeat signal, and extracting a heartbeat frequency. Through joint evaluation of the energy loss rate and the physiological frequency band energy ratio, the optimal balance parameter upper limit value can be adaptively searched according to individual characteristics, and the modal aliasing problem is effectively solved; according to the dynamic sparse reconstruction based on energy perception, the zero attraction intensity can be adjusted in real time according to spectrum energy distribution, and high-resolution and high-robustness heartbeat frequency estimation is realized under a short-time observation window.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of non-contact vital sign detection, and particularly relates to a FMCW radar vital sign detection method based on combination of SVMD and SSR. BACKGROUND

[0002] Traditional vital sign monitoring devices mainly rely on contact sensors such as patch electrodes, piezoelectric sensors, etc. Although these devices have high measurement accuracy, they need to directly contact the skin, which not only limits the freedom of user activities, but also may cause skin discomfort after long-term wearing, and is limited in application in special scenarios such as burn patient monitoring and infectious disease isolation.

[0003] Non-contact vital sign detection technology is increasingly valued for its convenience and comfort. This technology monitors vital signs through radar, infrared sensors or other wireless signals without the need for users to wear or directly contact any electronic devices. Among many non-contact sensing technologies, frequency modulated continuous wave (FMCW) radar has become a research hotspot in this field due to its high range resolution, low power consumption, good privacy protection and all-weather working advantages. FMCW radar transmits a linear frequency modulated signal and receives the echo reflected by the human chest wall. It uses the Doppler effect to extract the tiny phase changes caused by breathing and heartbeat, thereby realizing the extraction of vital signs.

[0004] However, in practical applications, the accurate detection of vital signs based on FMCW radar still faces severe challenges, mainly in the following two aspects.

[0005] Firstly, it is difficult to separate signals and resist interference. The human chest wall movement is the superposition of respiratory movement and heartbeat movement. Since the energy of the respiratory signal is much greater than that of the heartbeat signal, and the respiratory movement is usually non-sinusoidal, the high harmonic components generated by the respiratory movement are extremely easy to overlap with the weak heartbeat signal in the frequency domain, i.e. spectrum aliasing. In order to separate the respiratory and heartbeat signals, the existing signal decomposition algorithms such as empirical mode decomposition (EMD) lack a strict mathematical theoretical basis, and are prone to modal aliasing and end effect; although the variational mode decomposition (VMD) introduces a variational framework, its performance is highly dependent on the preset number of modes and the balance parameter. If the parameters are not properly set, it will lead to under-decomposition or over-decomposition of the signal. The successive variational mode decomposition (SVMD) avoids the preset number of modes, but the upper limit value of the balance parameter is usually a fixed empirical value. This parameter directly determines the final bandwidth constraint strength of the mode decomposition, and the fixed parameter is difficult to adapt to the differences of different individual physiological signals; the SVMD algorithm itself lacks a discrimination mechanism for the physical meaning of the signal, and cannot automatically identify and remove the respiratory harmonic interference mixed in the heartbeat mode, which limits the accuracy of heart rate estimation.

[0006] Secondly, the frequency estimation accuracy is insufficient under short-time observation. In order to meet the needs of real-time monitoring, the time window of radar data processing is usually short. According to the signal processing theory, the shorter the observation time, the lower the frequency resolution of Fourier transform, and there is serious spectrum leakage and fence effect, which leads to large heart rate estimation error. Although the algorithm based on sparse signal reconstruction (SSR) can break through the resolution limit, the existing algorithm usually uses a fixed sparse penalty term, which cannot adapt to the dynamic changes of radar echo signal-to-noise ratio. When the environmental noise is large or the target is micro-moving, the fixed penalty strength often leads to excessive suppression of the real signal or incomplete suppression of the noise, reducing the robustness of detection.

[0007] Therefore, it is necessary to provide an FMCW radar vital sign signal detection method which can adaptively optimize the decomposition parameters, actively suppress harmonic interference, and dynamically adjust the sparse reconstruction strategy according to the signal energy distribution, in order to solve the above problems. SUMMARY

[0008] To address the shortcomings of existing technologies, the technical problem this invention aims to solve is to provide an FMCW radar vital sign detection method based on a combination of SVMD and SSR. This method aims to solve the problems in existing technologies, such as the difficulty in extracting heartbeat signals due to respiratory harmonic interference, insufficient frequency resolution under short observation windows, and poor adaptability caused by fixed parameters in traditional sparse reconstruction algorithms. This will improve the accuracy and robustness of non-contact vital sign detection.

[0009] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A method for detecting vital signs using FMCW radar based on a combination of SVMD and SSR, characterized by the following steps: Step 1: Acquire FMCW radar echo signals, preprocess them, and obtain the micro-motion phase signals of the human chest wall; Step 2: The improved SVMD algorithm is used to decompose the micro-motion phase signal of the human chest wall, and the upper limit of the optimal equilibrium parameter is determined by the joint evaluation criteria to obtain the modal components. Set the search range for the upper limit of the equilibrium parameter, perform SVMD decomposition on each upper limit of the equilibrium parameter, and obtain a set of modal components respectively; calculate the energy loss rate corresponding to each upper limit of the equilibrium parameter according to equation (9). : (9) in, The total energy of all modal components. For the first The total energy of each modal component The number of modal components; Calculate the energy percentage of each modal component in the respiratory and cardiac frequency bands: (10) (11) in, For the first The energy of the respiratory signal in each modal component For the first The energy of the center jump signal of each modal component and The preset threshold; The upper limit of the balance parameter with the minimum energy loss rate and the energy proportion in the respiratory and cardiac frequency bands that meets the preset threshold is selected as the upper limit of the optimal balance parameter. The upper limit of the optimal balance parameter is used to decompose the micro-motion phase signal of the human chest wall. Step 3: Harmonic detection and signal reconstruction, reconstructing a pure heartbeat signal; Fourth step: build the SSR model, reconstruct the sparse spectrum based on the EB-ZA-SLMS algorithm, solve the pure heartbeat signal, and extract the heartbeat frequency.

[0010] Further, in the third step, all modal components are divided into a respiratory modal set and a heartbeat modal set according to the frequency, the frequency with the maximum energy in the respiratory modal set is selected as the respiratory fundamental frequency, all modal components in the heartbeat modal set are traversed, whether the center frequency of the modal component satisfies the frequency multiplication relationship is judged, if yes, the modal component is removed, otherwise, the modal component is reserved, and all reserved modal components are added to reconstruct the pure heartbeat signal.

[0011] Compared with the prior art, the present application has the beneficial effects that: The present application overcomes the limitations of traditional decomposition algorithms through joint evaluation of energy loss rate and physiological frequency band energy proportion, can adaptively find the upper limit value of the best balance parameter according to different individual characteristics, effectively solves the modal aliasing problem, introduces a harmonic detection and removal mechanism based on the respiratory fundamental frequency, actively identifies and filters out the high-energy respiratory harmonic interference in the signal reconstruction stage, significantly improves the signal-to-interference ratio of the weak heartbeat signal, and realizes high-resolution and high-robustness heartbeat frequency estimation under a short observation window, solves the contradiction between real-time performance and frequency resolution. BRIEF DESCRIPTION OF DRAWINGS

[0012] Figure 1 The present application is a whole flowchart; Figure 2 The present application is a flowchart of radar echo signal preprocessing and phase signal extraction; Figure 3 The present application is a flowchart of parameter adaptive optimization and harmonic removal of the improved SVMD algorithm; Figure 4 The present application is a flowchart of sparse spectrum reconstruction and heart rate extraction based on the EB-ZA-SLMS algorithm. DETAILED DESCRIPTION

[0013] The present application will be further described in detail below in combination with the drawings and specific embodiments.

[0014] As shown in the drawings, Figure 1 The present application provides a FMCW radar vital sign detection method based on the combination of SVMD and SSR, which comprises the following steps: First step: obtain the FMCW radar echo signal, preprocess it, and obtain the human chest wall micro-motion phase signal; As shown in the drawings, Figure 2As shown, first, the linear frequency modulation signal emitted by the FMCW radar is mixed with the echo signal and low-pass filtered to obtain an intermediate frequency signal ; (1) wherein, is the amplitude, is the beat frequency related to the distance, is the target distance, is the wavelength, is the imaginary part, and the phase term contains the chest wall micro-motion information of the human body; Next, the intermediate frequency signal is analog-to-digital converted, and the distance dimension fast Fourier transform is performed on the sampling data matrix to obtain a distance-slow time matrix, and the mean value subtraction method is used to filter out static clutter; (2) Finally, the human target is positioned by searching the distance bin with the maximum energy variance, and the phase signal at the position is extracted ; the phase is extracted using the arctangent, the phase jump is eliminated through the phase unwrapping algorithm, and then the differential processing is performed to remove the direct current drift, thereby obtaining the chest wall micro-motion phase signal of the human body .

[0015] Second step: the improved SVMD algorithm is used to decompose the chest wall micro-motion phase signal of the human body, the optimal balance parameter upper limit value is determined through the joint evaluation criterion, and a group of modal components are obtained; As shown in Figure 3 , the core idea of the SVMD algorithm is to convert the decomposition problem of a complex signal into a series of sequential variational optimization sub-problems. Unlike the traditional VMD algorithm which needs to pre-set the number of modes, the SVMD extracts the IMF with the most compact spectrum in the signal one by one through iteration until the residual signal meets a specific stopping condition, thereby realizing the adaptive determination of the number of modes.

[0016] In order to extract the first modal, the SVMD constructs the following constrained variational problem: (3) wherein, is the first modal, is the residual signal after the first decomposition; It is worth noting that in order to ensure the convergence of the decomposition, the SVMD considers that contains the first extracted modal and the remaining component that has not been decomposed , that is: (4) In order to ensure the accuracy of signal reconstruction while making the extracted first... Modality To achieve minimum bandwidth and avoid spectral overlap with extracted modes, SVMD constructs the following constrained variational optimization objective function: (5) The specific physical meanings of the three constraint terms are as follows: Classic Wiener filter constraints are used to minimize modes. Its bandwidth allows its spectral energy to closely revolve around the center frequency. distributed; Introducing the frequency response as The filter for the residual signal Filtering is performed to minimize the energy of the filtered residual signal, forcing the residual signal to... Minimize the number of nearby spectral components to ensure the integrity of the current mode extraction; : For the extracted previous Center frequency of the first mode Design a filter to prevent the currently extracted... It includes the spectral components of the extracted modes, thus avoiding mode aliasing. This optimization problem can be solved iteratively using the alternating direction multiplier method. Center frequency and Lagrange multiplier The iterative update formula for is as follows: (6) (7) (8) When the convergence condition is met, the current mode extraction is complete, and the algorithm proceeds to the next decomposition.

[0017] Since the upper limit of the equilibrium parameter directly affects the decomposition accuracy, this invention constructs a joint evaluation criterion for adaptive optimization; and sets the search range of the upper limit of the equilibrium parameter. Perform the above SVMD decomposition on each upper limit value of the equilibrium parameter to obtain a set of modes. ; Calculate the energy loss rate : (9) Simultaneously calculate the energy proportion of each mode in the respiratory frequency band (0.1-0.6Hz) and the heart rate band (0.8-2.0Hz): (10) (11) wherein, is the total energy of the th modal component, is the energy of the respiratory signal in the th modal component, is the energy of the heartbeat signal in the th modal component. and is a preset threshold value, which can be set according to actual requirements. Select the as the optimal parameter , and use the parameter to perform final decomposition on the signal.

[0018] Step 3: Harmonic detection and signal reconstruction, to obtain a pure heartbeat signal; Due to the high-order harmonics of the respiratory signal non-sinusoidal characteristics, which are easy to mix into the heartbeat mode, the modal obtained by decomposition is divided into a respiratory modal set and a heartbeat modal set according to the frequency, and the frequency with the maximum energy in the respiratory modal set is selected as the respiratory fundamental frequency ; all components in the heartbeat modal set are traversed to check whether the center frequency satisfies the frequency multiplication relationship ( ); if it satisfies, it is determined as respiratory harmonic interference, which is removed; the remaining heartbeat modal components after removing the interference are added to reconstruct a pure heartbeat signal .

[0019] Step 4: Constructing an SSR model, performing sparse spectrum reconstruction based on EB-ZA-SLMS to solve the heartbeat signal and extract the heartbeat frequency; As shown in Figure 4 , a frequency domain sparse representation model of the heartbeat signal is established: (12) wherein is the heartbeat signal sequence reconstructed in step S3, is a sparse spectrum vector to be reconstructed, is a noise vector, is an over-complete Fourier basis matrix, and the elements thereof are defined as: (13) The SSR model is mapped to an adaptive filtering process, wherein the sparse spectrum vector to be reconstructed corresponds to a weight vector of the adaptive filter, and the row vector of the basis matrix is mapped to an input vector and observe signal elements As expected response To suppress noise while protecting the weak heart rate, an energy sensing mechanism is introduced.

[0020] Define the energy percentage of the heartbeat spectrum : (14) in, The weight vector of the first One portion, A set of indices for heart rate ranges. Let the filter order be . To prevent tiny constants with a denominator of zero; Based on this, a dynamic zero attraction factor is constructed. : (15) In the formula, Based on the base penalty, To adjust the sensitivity factor. The physical meaning of this formula is: when the spectral energy is concentrated in the heartbeat frequency band, that is... When it is large, Automatically reduces compression of the true signal; when noise is high, Increase the size to enhance the sparsity constraint capability.

[0021] An improved signed least mean square algorithm is used for weight updates: (16) In the formula, For the input vector, For complex number sign functions, Step size factor The prediction error is expressed as: (17) After iterative convergence, The square of the modulus is the high-resolution heartbeat power spectrum, and the accurate heartbeat frequency can be obtained by searching for spectral peaks.

[0022] Any aspects not covered in this invention are applicable to existing technologies.

Claims

1. A method for detecting vital signs using FMCW radar based on a combination of SVMD and SSR, characterized in that, Includes the following steps: Step 1: Acquire FMCW radar echo signals, preprocess them, and obtain the micro-motion phase signals of the human chest wall; Step 2: The improved SVMD algorithm is used to decompose the micro-motion phase signal of the human chest wall, and the upper limit of the optimal equilibrium parameter is determined by the joint evaluation criteria to obtain the modal components. Set the search range for the upper limit of the equilibrium parameter, perform SVMD decomposition on each upper limit of the equilibrium parameter, and obtain a set of modal components respectively; calculate the energy loss rate corresponding to each upper limit of the equilibrium parameter according to equation (9). : （9） in, The total energy of all modal components. For the first The total energy of each modal component The number of modal components; Calculate the energy percentage of each modal component in the respiratory and cardiac frequency bands: （10） （11） in, For the first The energy of the respiratory signal in each modal component For the first The energy of the center jump signal of each modal component and The preset threshold; The upper limit of the balance parameter with the minimum energy loss rate and the energy proportion in the respiratory and cardiac frequency bands that meets the preset threshold is selected as the upper limit of the optimal balance parameter. The upper limit of the optimal balance parameter is used to decompose the micro-motion phase signal of the human chest wall. Step 3: Harmonic detection and signal reconstruction, reconstructing a pure heartbeat signal; Step 4: Construct an SSR model, perform sparse spectrum reconstruction based on the EB-ZA-SLMS algorithm, solve for the pure heartbeat signal, and extract the heartbeat frequency.

2. The FMCW radar vital sign detection method based on the combination of SVMD and SSR according to claim 1, characterized in that, In the third step, all modal components are divided into respiratory mode set and heartbeat mode set according to frequency, and the frequency with the largest energy in the respiratory mode set is selected as the respiratory fundamental frequency. Iterate through all modal components in the heartbeat modality set and determine whether the center frequencies of the modal components satisfy the harmonic relationship; If the condition is met, remove it; otherwise, keep it. By summing up all the retained modal components, a pure heartbeat signal can be reconstructed.

Images