Human body static target high-precision detection method and device based on radar life detector
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CENT SOUTH UNIV
- Filing Date
- 2026-06-03
- Publication Date
- 2026-06-30
Smart Images
Figure CN122307503A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data processing technology, and in particular to a high-precision detection method and device for static human targets based on a radar life detector. Background Technology
[0002] Ultra-wideband (UWB) radar, with its advantages of non-contact operation, strong penetration, resistance to harsh environments, and high resolution, has broad application value and prospects in fields such as post-disaster rubble rescue, life monitoring in confined spaces, and counter-terrorism and security. Among UWB radars, frequency-modulated continuous wave (FM-CW) UWB radar and frequency-stepped continuous wave (FROM-CW) UWB radar have become the most commonly used technologies in the UWB radar field due to their higher range resolution, better bandwidth expansion, and system stability. However, in the core application scenario of detecting stationary human targets, because the only movement of a stationary human target is the slight fluctuation of the chest cavity caused by breathing and heartbeat, the echo energy of vital signs is extremely weak and easily submerged by strong background clutter, system noise, and environmental interference. Existing technologies have a core deficiency: they cannot simultaneously ensure the quality of vital sign signal extraction and decomposition processing efficiency under conditions of strong clutter and low signal-to-noise ratio, making it difficult to achieve high-precision and rapid detection of stationary human targets. The reasons for this are as follows:
[0003] 1. The traditional robust principal component analysis (RPCA) algorithm based on the alternating direction multiplier method (ADMM) uses default settings for the core parameters of the sparse matrix iterative update formula for general scenarios. This cannot adapt to the narrow-band, weak, and non-stationary characteristics of radar vital signs signals, resulting in insufficient extraction accuracy of effective vital signs from sparse components. In low signal-to-noise ratio environments, the vital signs of stationary human targets are submerged by noise, and the detection success rate drops significantly.
[0004] 2. The core input parameters of the traditional RPCA algorithm, such as the penalty parameter and residual convergence threshold, have not been optimized for radar life detection scenarios. During the algorithm iteration process, the convergence efficiency of low-rank components and sparse components is low, requiring a large number of iterations to reach the convergence condition, which greatly increases the computation time and reduces the detection efficiency. It cannot meet the high-efficiency and real-time detection requirements of radar life detectors in on-site rescue scenarios. Summary of the Invention
[0005] This application proposes a high-precision detection method and device for static human targets based on a radar life detector, which can solve one of the problems existing in the background technology.
[0006] To achieve the above objectives, this application adopts the following technical solution:
[0007] Firstly, a high-precision method for detecting stationary human targets based on a radar life detector is provided, including:
[0008] After the radar life detector transmits the stepping sub-pulse signal, it receives the echo signal;
[0009] Perform a range-dimensional inverse Fourier transform on the echo to obtain a range-time matrix;
[0010] Furthermore, optimizing the solution of the distance-time matrix decomposition problem until convergence specifically involves: introducing dual variables to guide the alternating solution of subproblems, and iteratively updating the low-rank matrix and sparse matrix obtained by the distance-time matrix decomposition, wherein the low-rank matrix is used to characterize noise signals and the sparse matrix is used to characterize vital signs signals.
[0011] Based on the above technical solution, the radar is subjected to a fast-time inverse Fourier transform to convert the frequency axis into the range axis. The low-rank matrix and sparse matrix are updated iteratively, the dual variables are updated and the penalty parameters are modified. After convergence, the separation result of the target and the background is obtained. The background information is stored in the low-rank matrix, while the signal containing life feature information is stored in the sparse matrix, thereby realizing high-precision detection of static human targets.
[0012] In one possible design of the first aspect, the high-precision human body static target detection method based on radar life detector further includes:
[0013] The iterative update parameters are set, including: a sparse penalty coefficient for controlling the sparsity of the sparse matrix, a penalty parameter for balancing the convergence speed of the original residual and the dual residual, a convergence threshold for characterizing the iteration stopping condition, and a maximum number of iterations.
[0014] In one possible design approach of the first aspect, the decomposition problem of the distance-time matrix is:
[0015] Where X is the distance-time matrix, L is the low-rank matrix, and S is the sparse matrix. The sparsity penalty coefficient, Let L be the nuclear norm. Let S be the F-norm of S.
[0016] In one possible design approach of the first aspect, the update of the low-rank matrix involves: fixing the sparse matrix and the dual variable, and solving the nuclear norm minimization problem of the low-rank matrix, specifically including:
[0017] right Perform singular value decomposition:
[0018] The singular value matrix V contains elements less than The elements of the threshold are then subjected to soft thresholding to obtain a soft threshold matrix. ;
[0019] Update dual variable Y:
[0020] Where L is the left singular matrix r is The rank of V is the singular value matrix. , It is the transpose of a right singular matrix. K represents a slow time frame. It is a soft threshold matrix. It is a threshold parameter.
[0021] In one possible design approach of the first aspect, the sparse matrix is updated by fixing the low-rank matrix and the dual variable, and using soft thresholding to solve the sparsification subproblem of the L2 norm constraint of the sparse matrix.
[0022] In one possible design approach of the first aspect, optimizing the solution of the distance-time matrix decomposition problem further involves updating the dual variables and penalty parameters, specifically including:
[0023] First, calculate the residual. In order to judge The degree to which constraints are satisfied;
[0024] Next, update the dual variables. ;
[0025] Finally, adjust the penalty parameters. By increasing To accelerate convergence.
[0026] In one possible design approach of the first aspect, the relative error e is compared with the convergence threshold. The comparison results are used to determine whether convergence has occurred, and the relative error e is:
[0027] in, Let Z be the F-norm of the residual. Let X be the F-norm of the distance-time matrix X.
[0028] In one possible design of the first aspect, the high-precision human body static target detection method based on radar life detector further includes:
[0029] Before convergence, the distance-time matrix is preprocessed by iteratively updating the low-rank matrix, sparse matrix, and dual variables obtained from the distance-time matrix decomposition. The preprocessing includes direct-coupled wave calibration and amplitude normalization.
[0030] In one possible design of the first aspect, the direct-coupled wave calibration includes:
[0031] The position of the direct-coupled wave in the range-time matrix is determined by the energy corresponding to each range gate in the range-time matrix;
[0032] Furthermore, based on the position of the direct-coupled wave, a distance gate axis cyclic shift calibration is performed.
[0033] In a second aspect, an electronic device is provided, comprising: a processor and a memory coupled to the processor, the memory for storing a computer program; the processor for executing the computer program stored in the memory to cause the electronic device to perform the high-precision human static target detection method based on a radar life detector as described in any possible implementation of the first aspect. Attached Figure Description
[0034] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0035] Figure 1 This is a technical flowchart provided in the embodiments of this application;
[0036] Figure 2 This is a schematic diagram of direct-coupled wave calibration provided in an embodiment of this application;
[0037] Figure 3 This is a diagram showing the effect of the sparse matrix S after decomposition by the traditional ADMM-RPCA algorithm provided in the embodiments of this application;
[0038] Figure 4 This is the spectrum diagram of the sparse matrix S after decomposition by the traditional ADMM-RPCA algorithm, provided in the embodiments of this application;
[0039] Figure 5 This is a diagram showing the effect of the sparse matrix S after decomposition by the improved algorithm provided in the embodiments of this application;
[0040] Figure 6 This is the spectrum diagram of the sparse matrix S after decomposition by the improved algorithm provided in the embodiments of this application;
[0041] Figure 7 This is a comparison chart of the time-domain energy amplitudes of the traditional and improved target distance gate algorithms provided in the embodiments of this application;
[0042] Figure 8 This is a comparison chart of the frequency domain energy amplitude of the traditional and improved target distance gate algorithms provided in the embodiments of this application. Detailed Implementation
[0043] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0044] It should be noted that although functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification and the above-mentioned figures are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.
[0045] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.
[0046] This embodiment aims to develop an efficient decomposition method for high-precision extraction of radar vital signs signals, so as to achieve high-efficiency detection by radar life detectors.
[0047] like Figure 1 The diagram shown is a technical flowchart of this embodiment, which mainly includes radar echo preprocessing and alternating direction multiplier method-robust principal component analysis (ADMM-RPCA) decomposition. The following is a detailed description of the technical solution of this embodiment.
[0048] 1. Radar life detector signal model:
[0049] The stepped-frequency ultra-wideband radar transmits N finely tuned frequency stepped sub-pulse signals within one pulse repetition period. It can be represented as: (1)
[0050] in This represents the energy of the transmitted signal, and N represents the number of frequency points. Represents a refined signal frequency. This indicates the initial frequency of the step-frequency radar. Indicates the frequency step size. Indicates the initial phase of the frequency point, and the total synthesized bandwidth of the radar. Distance resolution for t is the time, c is the speed of light, and n is the frequency.
[0051] When the signal encounters the target, it is modulated with the target signal and reflected, then received by the radar receiving antenna. At this point, the signal is received at the nth frequency. It can be represented as: (2)
[0052] in, It is the delay of the echo signal, which depends on the distance between the target and the radar.
[0053] 2. Inverse Fourier Transform in the Distance Dimension:
[0054] Performing an inverse Fourier transform on all N frequency points of a pulse repetition period can transform the frequency axis into the distance axis. A signal from the distance gate It can be represented as: (3)
[0055] This allows radar echoes to be converted into a range-time matrix.
[0056] 3. Distance-time matrix preprocessing:
[0057] (1) Direct-coupled wave calibration:
[0058] The radar's direct-coupled wave should ideally be at the first range gate, but due to system delay offsets, its position may be cyclically shifted to other locations. It needs to be calibrated to the correct position through cyclic shifting to eliminate system errors. First, the radar's direct-coupled wave needs to be located. Since the direct-coupled wave is the signal directly coupled to the receiving antenna after radar transmission, it is usually the range gate with the highest energy. Therefore, finding the range gate with the highest energy determines the direct-coupled wave range gate. Next, cyclic shifting calibration is performed, cyclically shifting the range gate axis until the direct-coupled wave range gate is located at the first range gate position. Assuming a signal frame has M range gates, and the P-th range gate has the highest energy, the direct-coupled wave calibration process is as follows: Figure 2 As shown:
[0059] (2) Amplitude normalization:
[0060] To eliminate the influence of the absolute magnitude of the signal amplitude, focus on the relative energy distribution, and facilitate subsequent threshold detection and visualization, the echo signal also needs to be normalized.
[0061] 4. Set ADMM-RPCA parameters:
[0062] Before applying the ADMM-RPCA algorithm to the preprocessed distance-time matrix, the algorithm parameters need to be initialized.
[0063] Sparsity penalty coefficient It controls the sparsity of the sparse matrix S. The larger the value, the sparser the S-shape. The smaller the value, the more noise is incorporated into S. For a matrix with M distance gates and a slow time axis of K frames... , The optimal value is .
[0064] Penalty parameters It is the core penalty factor of ADMM, used to balance the convergence speed of the original residual and the dual residual. It can be understood as the step size of the iteration. The commonly used theoretical value is , It is the F-norm of X, equal to the square root of the sum of the squares of all its elements. However, when using radar to detect vital signs of stationary human targets, the penalty parameter can be appropriately increased to... This allows both the dual residual and the original residual to converge quickly, improving the real-time performance of the algorithm.
[0065] Convergence threshold This is the iteration stopping condition; the iteration terminates when the residual is less than a threshold. Setting it too high will result in excessively large residuals, leading to poor performance; setting it too low will affect the number of iterations, potentially causing a prolonged iteration without yielding satisfactory results. Therefore, we use [the threshold here]. .
[0066] Maximum number of iterations This is the maximum number of iterations set to prevent iteration divergence or excessive algorithm runtime. Because the algorithm has good performance and can iterate to the result quickly, it is set to 1000 here, but in most cases this number of iterations will not be reached.
[0067] 5. ADMM-RPCA Iterative Algorithm:
[0068] The goal of the ADMM-RPCA algorithm is to iteratively update the low-rank matrix L, the sparse matrix S, and the dual variable Y, while dynamically adjusting the penalty parameter. This achieves accurate decomposition of low-rank sparse matrices. Because background clutter is often smooth and redundant, its matrix should be low-rank, while the matrix containing human vital signs signals has few abrupt changes and is a sparse matrix. Therefore, the ADMM-RPCA algorithm separates the target from the background in data mixed with noise and clutter, dividing the original echo into a low-rank matrix L storing background information and a sparse matrix S storing the target's vital signs signals. Decomposing the image matrix essentially involves... Solving optimization problems under certain conditions , The nuclear norm of L is equal to the sum of all its singular values. Let S be the F-norm of S. However, directly solving this constrained optimization problem is very difficult. Therefore, it is necessary to divide the problem into two subproblems, solve them alternately, and iterate to gradually approach the optimal solution. Thus, a dual variable Y needs to be introduced, and the iteration is divided into updating the low-rank matrix L, updating the sparse matrix S, updating the dual variable Y, and adjusting the penalty parameter. These three steps.
[0069] 5.1 Matrix Initialization:
[0070] First establish The empty matrix, low-rank matrix L, sparse matrix S, and dual variable Y.
[0071] 5.2 Low-rank matrix update:
[0072] Given a fixed sparse matrix S and dual variable Y, solve the subproblem of the low-rank matrix L, namely the nuclear norm minimization problem, using the singular value thresholding method. First, [the following steps are taken]. Perform SVD singular value decomposition. (4)
[0073] Where L is the left singular matrix ( r is The rank of V is the singular value matrix (V is the rank of V). ), It is the transpose of a right singular matrix ( Thus, V is a diagonal matrix, with the elements on the diagonal arranged from largest to smallest.
[0074] Next, the value less than in V All elements of the threshold are set to zero, i.e., soft thresholding is performed, resulting in... This is equivalent to Feature values less than the threshold are set to zero. This differs from the threshold parameter commonly used in traditional ADMM algorithms. In contrast, this approach will miss more smaller feature values, and the vital signs information contained in these elements will be more easily preserved in the sparse matrix S.
[0075] Finally, update the dual variable Y: (5)
[0076] 5.3 Sparse Matrix Update:
[0077] With a fixed low-rank matrix L and dual variable Y, solve the subproblem of the sparse matrix S, namely, sparsification under L2 norm constraints, using the soft thresholding method. Perform soft thresholding to reduce values less than the threshold. All elements are set to zero, resulting in the updated sparse matrix S.
[0078] Finally, update the sparse matrix S: (6)
[0079] Where sign is the sign function. It is the element-wise product of matrices.
[0080] 5.4 Update of Dual Variables and Penalty Parameters:
[0081] The purpose of this step is to modify the dual variable and increase its size. To accelerate convergence, the residual Z is calculated until it is sufficiently small. First, the residual Z is calculated to determine... The degree to which constraints are satisfied. (7)
[0082] Next, update the dual variable Y: (8)
[0083] Finally, adjust the penalty parameters. We can accelerate convergence by increasing u, while setting an upper limit of 1e6 to prevent it from becoming too large. (9)
[0084] 5.5 Convergence Judgment:
[0085] First, calculate the current relative error. It can be expressed as the ratio of the F-norm of the residual Z to the F-norm of the original matrix X: (10)
[0086] Compare the relative error e with the set convergence threshold. If a comparison is made, If convergence has been achieved, output the final result: the low-rank matrix L and the sparse matrix S. Otherwise, repeat steps 5.2, 5.3, 5.4, and 5.5 until the convergence condition is met or the number of iterations exceeds the maximum number of iterations. .
[0087] To further demonstrate the effectiveness of this embodiment, the following comparison of the performance of the traditional ADMM-RPCA algorithm and the improved algorithm is conducted using actual test data, and the results are as follows. Figure 3 and Figure 4 The images show the effect diagram and corresponding spectrum of the sparse matrix S after decomposition by the traditional ADMM-RPCA algorithm. Figure 5 and Figure 6The images show the effect diagram and corresponding spectrum of the sparse matrix S after decomposition by the improved algorithm. It can be seen that the improved algorithm retains more energy near the target's distance gate, exhibiting more characteristics similar to vital signs such as breathing. Furthermore, the spectrum shows that since the human respiratory frequency is approximately between 0.2-0.5 Hz, the improved algorithm exhibits stronger energy near the 0.5 Hz frequency of the target's distance gate, proving that the improved algorithm retains more vital sign information. Figure 7 Comparing the time-domain energy amplitudes of the traditional and improved algorithms for the target distance gate, it can be seen that after normalizing the original echo energy, the improved algorithm retains more energy in each breathing cycle. Figure 8 Comparing the frequency domain energy amplitudes of the traditional and improved target range gate algorithms, the improved algorithm exhibits a larger energy peak around 0.5 Hz, preserving more energy from vital signs such as respiration. Furthermore, when both algorithms achieve the same level of low-rank partial error, the traditional algorithm requires 64 iterations, while the improved ADMM-RPCA only needs 7 iterations to separate the background from the target, demonstrating its superior real-time performance.
[0088] Key points of this embodiment:
[0089] (1) This embodiment proposes a high-precision detection method for static human targets based on radar life detector. The overall technical solution is divided into three core stages: radar echo data preprocessing, low-rank sparse decomposition based on optimized ADMM-RPCA, and extraction of vital signs signals and target determination. Based on the distribution model of radar vital signs signals, the ADMM-RPCA algorithm is systematically optimized, realizing high-precision and high-efficiency detection of static human targets in strong clutter and low signal-to-noise ratio environments.
[0090] (2) This embodiment has made core optimizations to the sparse matrix iterative update formula of the ADMM-RPCA algorithm, and changed the coefficient of the sparse matrix soft threshold shrinkage function in the traditional algorithm from ( ) optimized to ( It accurately adapts to the sparse distribution characteristics of radar vital signs signals, significantly improving the extraction accuracy of weak vital signs signals.
[0091] (3) In this embodiment, the core input parameters of the ADMM-RPCA algorithm are optimized for radar life detection scenarios, and the initial penalty parameter is increased. This increased the residual convergence threshold. While ensuring decomposition accuracy, it significantly accelerated the algorithm's iteration convergence speed and improved the efficiency of life signal decomposition and processing.
[0092] This embodiment proposes a high-precision detection method for static human targets based on a radar life detector, aiming to improve the extraction quality and decomposition efficiency of radar vital sign signals. The method mainly consists of two stages: radar raw echo preprocessing and ADMM-RPCA decomposition. In the radar raw echo preprocessing stage, an inverse Fourier transform is first performed on the radar fast time matrix to convert the frequency axis into the range axis. Secondly, the range-time matrix is aligned with the direct-coupled wave and normalized. In the ADMM-RPCA decomposition stage, optimized initial parameters are first set to ensure faster iteration. Then, an iterative loop begins, updating the low-rank matrix and sparse matrix respectively. The residuals of the current low-rank matrix, sparse matrix, and the unseparated matrix are calculated, and the dual variables are updated and the penalty parameters are modified. Finally, the residuals of the decomposed matrix are compared to see if they are less than the convergence threshold or if the number of iterations is greater than the maximum number of iterations. If both conditions are met, the iteration exits and the decomposed matrix is output; otherwise, the iteration continues. The final result is the separation of the target and background. Background information is stored in the low-rank matrix, while signals containing vital sign information are stored in the sparse matrix. This invention combines the characteristics of vital signs signals of static human targets, modifies and improves the ADMM iterative formula and initial parameter settings, reduces the number of iterations while improving the retention of vital sign information in the sparse matrix.
[0093] This application also provides an electronic device, including: a processor, and a memory coupled to the processor, the memory being used to store a computer program; the processor being used to execute the computer program stored in the memory, so that the electronic device performs the method as described in any of the above embodiments.
[0094] Electronic devices can be computing devices such as desktop computers, laptops, handheld computers, and cloud servers. These electronic devices may include, but are not limited to, processors and memory.
[0095] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the electronic device, connecting various parts of the device via various interfaces and lines.
[0096] The memory can be used to store the computer program, and the processor implements various functions of the electronic device by running or executing the computer program stored in the memory and calling the data stored in the memory.
[0097] The memory may primarily include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.
[0098] This application also provides a storage medium, which is a computer-readable storage medium. The computer program is stored in the computer-readable storage medium, and when executed by a processor, the computer program can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.
[0099] This application also provides a computer program product, including: a computer program or instructions that, when the computer program or instructions are run on a computer, cause the computer to perform any of the above possible implementation methods.
[0100] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.
Claims
1. A high-precision detection method for stationary human targets based on a radar life detector, characterized in that, include: After the radar life detector transmits the stepping sub-pulse signal, it receives the echo signal; Perform a range-dimensional inverse Fourier transform on the echo to obtain a range-time matrix; Furthermore, optimizing the solution of the distance-time matrix decomposition problem until convergence specifically involves: introducing dual variables to guide the alternating solution of subproblems, and iteratively updating the low-rank matrix and sparse matrix obtained by the distance-time matrix decomposition, wherein the low-rank matrix is used to characterize noise signals and the sparse matrix is used to characterize vital signs signals.
2. The high-precision detection method for static human targets based on radar life detector as described in claim 1, characterized in that, The high-precision detection method for stationary human targets based on radar life detectors also includes: The iterative update parameters are set, including: a sparse penalty coefficient for controlling the sparsity of the sparse matrix, a penalty parameter for balancing the convergence speed of the original residual and the dual residual, a convergence threshold for characterizing the iteration stopping condition, and a maximum number of iterations.
3. The high-precision detection method for static human targets based on radar life detector as described in claim 1, characterized in that, The decomposition problem of the distance-time matrix is as follows: Where X is the distance-time matrix, L is the low-rank matrix, and S is the sparse matrix. The sparsity penalty coefficient, Let L be the nuclear norm. Let S be the F-norm of S.
4. The high-precision detection method for stationary human targets based on a radar life detector as described in claim 3, characterized in that, The update of the low-rank matrix involves fixing the sparse matrix and the dual variable, and solving the nuclear norm minimization problem of the low-rank matrix, specifically including: right Perform singular value decomposition: The singular value matrix V contains elements less than The elements of the threshold are then subjected to soft thresholding to obtain a soft threshold matrix. ; Update dual variable Y: Where L is the left singular matrix r is The rank of V is the singular value matrix. , It is the transpose of a right singular matrix. K represents a slow time frame. It is a soft threshold matrix. It is a threshold parameter.
5. The high-precision detection method for static human targets based on a radar life detector as described in claim 4, characterized in that, The update of the sparse matrix is as follows: fix the low-rank matrix and the dual variable, and use soft thresholding to solve the sparsification subproblem of the L2 norm constraint of the sparse matrix.
6. The high-precision detection method for static human targets based on radar life detector as described in claim 5, characterized in that, Optimizing the solution to the distance-time matrix decomposition problem also involves updating the dual variables and penalty parameters, specifically including: First, calculate the residual. In order to judge The degree to which constraints are satisfied; Next, update the dual variables. ; Finally, adjust the penalty parameters. By increasing To accelerate convergence.
7. The high-precision detection method for static human targets based on radar life detector as described in claim 2, characterized in that, Using the relative error e and the convergence threshold The comparison results are used to determine whether convergence has occurred, and the relative error e is: in, Let Z be the F-norm of the residual. Let X be the F-norm of the distance-time matrix X.
8. The high-precision detection method for static human targets based on radar life detector as described in claim 1, characterized in that, The high-precision detection method for stationary human targets based on radar life detectors also includes: Before convergence, the distance-time matrix is preprocessed by iteratively updating the low-rank matrix, sparse matrix, and dual variables obtained from the distance-time matrix decomposition. The preprocessing includes direct-coupled wave calibration and amplitude normalization.
9. The high-precision detection method for static human targets based on a radar life detector as described in claim 8, characterized in that, The direct-coupled wave calibration includes: The position of the direct-coupled wave in the range-time matrix is determined by the energy corresponding to each range gate in the range-time matrix; Furthermore, based on the position of the direct-coupled wave, a distance gate axis cyclic shift calibration is performed.
10. An electronic device, characterized in that, The electronic device includes: a processor, and a memory coupled to the processor. The memory is used to store computer programs; The processor is configured to execute the computer program stored in the memory, so that the electronic device performs the high-precision detection method for static human targets based on a radar life detector as described in any one of claims 1-9.