Research On DOA Estimation Algorithm Based On Sparse Reconstruction Theory

As an significant part of modern signal processing,underwater acoustic array signal processing is widely used in target detection,underwater acoustic communication,ship navigation,underwater image processing and other fields.This paper is oriented to the low signal-to-noise ratio,small snapshots,and narrow-band signal DOA estimation requirements in the underwater acoustic field.Using the sparsity of the signal space,the on-grid model,off-grid model and grid-less model are established,furthermore,high-precision DOA estimation is completed based on the sparse reconstruction theory.First,starting from the array element domain data,establish on-grid and off-grid models,and study sparse reconstruction methods,including Sparse Bayesian Learning(SBL),Off-grid Sparse Bayesian Inference(OGSBI),Perturbed Sparse Bayesian Learning(PSBL)and Root Sparse Bayesian Learning(Root SBL),and Sparse Bayesian Algorithm Based on Steepest Descent Method(SDSBL)is proposed to realize the narrow-band independent signal DOA estimation.Next,three methods of SVD decomposition,Unitary Rransformation,and Dynamic Threshold Setting are introduced to achieve high-efficiency DOA estimation.Then a high-precision DOA estimation framework based on signal spatial feature reconstruction is proposed to realize the joint estimation of the number of sources,noise,signal power and DOA.The simulation results show that,compared with the subspace method,the sparse reconstruction method has better low signal-to-noise ratio,small snapshot adaptability,higher DOA estimation accuracy,and the SDSBL algorithm has a smaller amount of calculation;efficient DOA Although the estimation method reduces the amount of calculation,the scope of application is narrow;the sparse reconstruction algorithm under the framework of high-precision DOA estimation based on signal spatial feature reconstruction is effective and predominant in low signal-to-noise ratio,small snapshot,correlation and even coherent signal scenarios,in addition,it also can provide more effective prior knowledge for target detection and tracking.Next,the array element domain data is extended to the covariance vector to improve the adaptability of each algorithm in low-noise ratio scenarios,while reducing the amount of calculation,and using the Toeplitz characteristic of the ULA covariance matrix to further reduce the amount of calculation.For narrow-band correlation signals,the correlation between different covariance vectors is used to estimate the error,and the decorrelation processing is realized without losing the array aperture.In addition,to solve the problem of low estimation accuracy of the covariance matrix of small snapshot scenes,a covariance fitting standard that is not sensitive to related signals is introduced,with the help of SParse Iterative Covariance-based Estimation(SPICE)and Sparse and Parametric Appraoch(SPA),establish a grid-less model,and propose a grid-less q-SPICE algorithm(q-SPICE-PP).The simulation results show that,compared to the sparse reconstruction methods in the array element domain,the DOA estimation algorithm based on the covariance vector has better low signal-to-noise ratio adaptability,and has a lower computational complexity under the premise of ensuring the accuracy of DOA estimation.,But its performance depends on the estimation accuracy of the covariance matrix;the algorithm based on the covariance fitting standard has better DOA estimation accuracy in small snapshots,correlation and even coherent signal scenarios,but its computational complexity is larger.Finally,the validity and practicability of the algorithm proposed in this paper are proved through the lake test data processing.