CN-122017753-A - Radar-oriented order constraint jacobian feature value self-ordering method and system

CN122017753ACN 122017753 ACN122017753 ACN 122017753ACN-122017753-A

Abstract

The invention discloses a radar-oriented order constraint jacobian eigenvalue self-ordering method and a system, which solve the problem of resource consumption caused by an additional ordering module in the jacobian iterative eigenvalue decomposition post-processing flow in the prior art, realize the direct output of ordered eigenvalues and eigenvectors, and omit an independent ordering module; the method comprises the steps of processing a real symmetric matrix A obtained by radar baseband signal sampling and covariance estimation to obtain a compressed array, initializing a feature vector matrix V, generating index pairs according to a scheduling sequence and extracting submatrices in a parallel iteration process, generating two types of asymmetric rotation operators through a conditional drive mechanism, synchronously updating the feature vector matrix to track feature vectors, carrying out data global replacement, calculating off-diagonal element energy judgment convergence, directly extracting feature values arranged in descending order from main diagonal lines of the final compressed array after iteration is ended, and outputting corresponding feature vectors for estimating the direction of arrival.

Inventors

CAO YUNHE
TANG CHENGQIAN
LU YAXUAN
MEI LIRONG
LIANG BOWEN

Assignees

西安电子科技大学
中国电子科技集团公司第五十四研究所

Dates

Publication Date: 20260512
Application Date: 20260112

Claims (9)

1. The radar-oriented sequence constraint jacobian feature value self-ordering method is characterized by comprising the following steps of: S1, receiving a baseband signal acquired by a radar array, sampling and covariance estimating the baseband signal to obtain an n-order real-symmetry radar signal target matrix A to be decomposed, preprocessing a data structure of the radar signal target matrix A to obtain a radar compression array, and initializing an n-order eigenvector matrix V; S2, generating a plurality of column index pairs to be processed in the current parallel iteration period according to a preset scheduling sequence, and extracting diagonal submatrices from the radar compression array according to indexes of each index pair And according to the diagonal submatrix Calculating to obtain a first type rotation operator corresponding to the current parallel iteration period And a second class rotation operator ; S3, using the first type rotation operator For the diagonal submatrices Contract transformation is carried out to obtain an updated radar compression array, and a second class rotation operator is simultaneously used Performing similarity transformation on the corresponding columns of the feature vector matrix V to obtain updated feature vector columns, and further obtaining an updated feature vector matrix V; S4, after the updating of the current parallel iteration period is completed, carrying out global replacement on data in the updated radar compression array according to a preset scheduling rule to obtain a replaced radar compression array, judging whether the replaced radar compression array meets the iteration termination condition, if not, returning to the step S2 for the next iteration based on the replaced radar compression array and the updated eigenvector matrix V, and if so, entering the step S5; s5, extracting main diagonal elements from the final radar compression array to obtain n eigenvalues arranged in a descending order, and outputting a final eigenvector matrix V as a corresponding eigenvector for radar signal arrival direction estimation.
2. The radar-oriented sequence constraint jacobian feature value self-ordering method of claim 1, wherein the performing data structure preprocessing on the radar signal target matrix a to obtain a radar compression array comprises: extracting an upper triangular part of the symmetric matrix by utilizing the symmetry of the symmetric matrix to obtain an upper triangular matrix; and storing the upper triangular matrix into a one-dimensional array to obtain a radar compression array.
3. The method of claim 1, wherein the step of determining the order of the radar-oriented order-constrained jacobian eigenvalues is based on the diagonal submatrices Calculating to obtain a first type rotation operator corresponding to the current parallel iteration period And a second class rotation operator Comprising: calculating a main rotation angle for eliminating the diagonal sub-matrix off-diagonal elements And rotating the main rotation angle Conversion to main rotation parameters ; Comparing the diagonal submatrices First diagonal element of (a) And a second diagonal element Is of a size of (2); if the first diagonal element Greater than or equal to the second diagonal element Then the first class of rotation operators And a second class rotation operator Are all equal to the main rotation parameter ; If the first diagonal element Smaller than the second diagonal element Then for the main rotation parameter Applying sequence adjustment transformation and according to the transformed main rotation parameters Obtaining a first class of rotation operators And a second class rotation operator 。
4. The method of claim 3, wherein the main rotation angle is a self-ordering of the radar-oriented sequence-constrained jacobian eigenvalues The calculation formula of (2) is expressed as: ; Wherein, the Representing a first diagonal element; Representing a second diagonal element; representing a first off-diagonal element; Representing a second off-diagonal element; Representing an arctangent function.
5. The method for self-ordering radar-oriented order constraint jacobian eigenvalues of claim 3, characterized in that the pair of main rotation parameters Applying sequence adjustment transformation and according to the transformed main rotation parameters Obtaining a first class of rotation operators And a second class rotation operator Comprising: The main rotation parameter is set Applying an order adjustment operator to obtain a correction matrix ; According to the correction matrix For the main rotation parameter Obtaining a first class of rotation operators And a second class rotation operator 。
6. The method for self-ordering radar-oriented order constraint jacobian feature values of claim 1, wherein said using said first class rotation operator For the diagonal submatrices Contract transformation is carried out to obtain an updated radar compression array, which comprises the following steps: Using the first class rotation operator Constructing a rotation matrix According to the rotation matrix For the diagonal submatrices Performing orthogonal rotation transformation to eliminate the diagonal submatrices Off-diagonal elements in (a) and updating the diagonal submatrices Diagonal elements in (a) to obtain updated diagonal submatrices ; The updated diagonal submatrix And writing back to the radar compression array to obtain an updated radar compression array.
7. The method for self-ordering radar-oriented order constraint jacobian eigenvalues according to claim 1, wherein said rotation operators according to a second class Performing similarity transformation on the corresponding column of the feature vector matrix V to obtain an updated feature vector column, and further obtaining an updated feature vector matrix V, including: based on the second class rotation operator Performing similarity transformation on two columns corresponding to the current column index pair in the feature vector matrix V; and writing the two columns of vectors obtained after transformation back to the feature vector matrix V as updated feature vector columns to obtain the updated feature vector matrix V.
8. The radar-oriented order constraint jacobian feature value self-ordering method of claim 1, wherein the performing global permutation on the data in the updated radar compression array according to a preset scheduling rule to obtain a permuted radar compression array comprises: Remapping element indexes in the updated radar compression array through a fixed replacement network according to the cyclic scheduling sequence to generate a new index mapping relation; And rearranging the data in the updated radar compression array by utilizing the new index mapping relation to obtain a replaced radar compression array of which the data distribution meets the scheduling requirement of the next parallel iteration period.
9. A radar-oriented order-constrained jacobian feature value self-ordering system, comprising: The data preprocessing module is used for receiving baseband signals acquired by the radar array, sampling and covariance estimating the baseband signals to obtain an n-order real symmetric radar signal target matrix A to be decomposed, preprocessing a data structure of the radar signal target matrix A to obtain a radar compression array, and initializing an n-order eigenvector matrix V; an operator calculation module for generating a plurality of column index pairs to be processed in the current parallel iteration period according to a preset scheduling sequence, and extracting diagonal submatrices from the radar compression array according to indexes of each index pair And according to the diagonal submatrix Calculating to obtain a first type rotation operator corresponding to the current parallel iteration period And a second class rotation operator ; A parallel processing module for using the first type rotation operator For the diagonal submatrices Contract transformation is carried out to obtain an updated radar compression array, and a second class rotation operator is simultaneously used Performing similarity transformation on the corresponding columns of the feature vector matrix V to obtain updated feature vector columns, and further obtaining an updated feature vector matrix V; The global replacement module is used for carrying out global replacement on the data in the updated radar compression array according to a preset scheduling rule after completing the updating of the current parallel iteration period, obtaining a replaced radar compression array, judging whether the replaced radar compression array meets the iteration termination condition, if not, returning to the operator calculation module for carrying out the next iteration based on the replaced radar compression array and the updated eigenvector matrix V, and if so, entering the output module; The output module is used for extracting main diagonal elements from the final radar compression array to obtain n eigenvalues arranged in descending order, and outputting a final eigenvector matrix V as a corresponding eigenvector for radar signal arrival direction estimation.

Description

Radar-oriented order constraint jacobian feature value self-ordering method and system Technical Field The invention relates to the technical field of radar signal processing, in particular to a radar-oriented sequence constraint jacobian eigenvalue self-ordering method and system. Background In the field of radar signal processing, high-precision direction of arrival (DOA) estimation is one of the core tasks. The performance of classic super-resolution DOA estimation algorithms such as MUSIC depends on accurate and fast eigenvalue decomposition (EVD) of the antenna array covariance matrix. For modern radar systems such as phased array radar, the processing of a plurality of snapshot data needs to be completed within a very short time window, so that the EVD operation must be completed within a microsecond level, which puts a severe requirement on the real-time performance of computing hardware. In terms of hardware platform selection, the FPGA becomes a mainstream platform for realizing an EVD module in a radar real-time signal processing system due to a customizable parallel computing architecture, deterministic low processing delay and efficient pipeline processing capacity. In an EVD implementation scheme based on an FPGA, a parallel Jacobi (Jacobi) iterative method is widely adopted because of its regular operation structure and natural suitability for hardware parallelization. The existing common parallel Jacobi scheme generally comprises the following procedures of dividing a symmetric matrix into a plurality of 2×2 submatrices for parallel processing, calculating a rotation operator and applying the rotation operator to the submatrices to realize diagonalization, and iterating until the energy of non-diagonal elements of the matrix is converged below a threshold value. However, the prior art scheme has an inherent defect that the real-time performance of the radar system is severely limited, and since the core goal of the Jacobi method is only diagonalization, the iterative process of the Jacobi method does not have a mechanism for ordering the characteristic values. Therefore, after iteration converges, the obtained eigenvalues and corresponding eigenvectors are arranged unordered when output. In radar DOA estimation applications, it is necessary to distinguish signal subspaces from noise subspaces according to the magnitudes of the eigenvalues, typically in descending order, and select the largest number of eigenvalues and their corresponding eigenvectors for subsequent spatial spectrum estimation. For this reason, existing solutions have to concatenate a separate ordering module after the EVD calculation module. The sorting module needs to sort the eigenvalue sequences and synchronously adjust the column sequence of the eigenvector matrix. The post-sequencing module has the obvious problems that firstly, extra FPGA logic resources such as comparators, multiplexers and complex state control logic are required to be consumed, secondly, sequencing operation and an EVD iteration process are in serial dependency relationship and cannot be subjected to pipeline optimization, so that non-negligible extra processing delay is introduced, and finally, the resources and time sequence cost of the whole EVD processing link are increased, and the bottleneck for improving the real-time performance of radar signal processing is formed. Therefore, the resource and delay expense brought by the independent sequencing module needs to be eliminated, and the urgent requirement of the radar system on high-real-time signal processing is met. Disclosure of Invention The invention solves the problem of resource consumption caused by an additional sorting module in the analysis post-processing flow of the jacobian iterative eigenvalue in the prior art by providing the self-sorting method and the system of the sequence constraint jacobian eigenvalue for the radar, and realizes that the EVD module directly outputs the ordered eigenvalue and the corresponding eigenvector, thereby omitting an independent sorting module. In a first aspect, the present invention provides a radar-oriented self-ordering method for sequence-constrained jacobian feature values, the method comprising: S1, receiving a baseband signal acquired by a radar array, sampling and covariance estimating the baseband signal to obtain an n-order real-symmetry radar signal target matrix A to be decomposed, preprocessing a data structure of the radar signal target matrix A to obtain a radar compression array, and initializing an n-order eigenvector matrix V; S2, generating a plurality of column index pairs to be processed in the current parallel iteration period according to a preset scheduling sequence, and extracting diagonal submatrices from the radar compression array according to indexes of each index pair And according to the diagonal submatrixCalculating to obtain a first type rotation operator corresponding to the current parallel iteration periodAnd a sec