Research On Multi-target DOA Estimation Based On L-shape Array

Direction-of-arrival(DOA)estimation is a basic problem in the field of array signal processing.It has obtained increasing attention in a wide field of wireless communication,radar,sonar,seismic exploration and so on.In the actual scene,the two-dimensional(2D)DOA estimation can better characterize the DOA of the target signal in the three-dimensional space.Many antenna geometries for 2D-DOA estimation have been proposed,such as L-shape array,uniform circular array,parallel linear array,rectangular array,etc.Among these array geometries,2D-DOA estimation with L-shape array has gained its popularity for its high estimation accuracy,moderate complexity,and simple design.Based on the L-shape array,this paper fully exploits the sparse characteristics of the target DOA in the two-dimensional angle domain,then establishes a more accurate L-shape array model and studies the 2D-DOA estimation methods to obtain outstanding performance.In order to overcome the fitting error of the signal model caused by the discrete angle domain,this paper studies the L-shape array DOA estimation methods based on the sparse Bayesian learning and the sparse gridless.The main research results of the paper are as follows:1.Aiming at the problem of mismatching dictionary bases of 2D-DOA estimation array model based on sparse representation under L-shaped array,we propose a grid adaptive model to alleviate model error caused by uniform discrete grid.We also propose a grid adaptive sparse Bayesian learning(GASBL)method for 2D-DOA estimation method with L-shape array.This method realizes the estimation of elevation and azimuth by sparse Bayesian learning framework,and completes the angle pair automatically.Simulation results illustrate that our approach can achieve a better performance than the state-of-the-art methods.2.For the model error problem caused by the discretization of the angle domain this paper studies sparse gridless methods which can directly estimate DOAs in the continuous angle domain.In order to overcome the model fitting error,the 2D-DOA is regarded as a sparse parameter in the continuous space of the 2D angular domain.We establish an array model represented by sparse parameters for L-shape array.Then we propose a gridless sparse method named CCM-SPA which is based on cross-correlation matrix and Vandermonde decomposition.This method first estimates the DOA of the sub-array separately,and then match obtained elevation and azimuth according to the cross-correlation matrix.Simulation results show that the method proposed in this paper can achieve good performance,which is not only suitable for uniform L-shape array,but also for sparse L-shape array.