Research On Algorithm Of Sparse DOA Estimation With Matrix Filter Under Strong Interference Environment

Under the strong interference environment,such radar,sonar,etc.,the received signals of the array contain both weak expected sources and strong interferences.Weak expected sources will be severely interfered or even drowned,which makes it impossible to accurately locate them.Therefore,the study of DOA estimation under the strong interference environment is particularly important.Although existing sparse algorithms under the strong interference environment can achieve great performance under certain conditions,the estimation accuracy of off-grid expected sources is low.Besides,the computational complexity of existing algorithms is too high,which is not applicable to practical situations.Aiming at the problems of existing algorithms,this paper makes research on the problem of sparse DOA estimation in a strong interference environment.The main work and contributions of this article are as follows:1.The two existing algorithms under the strong interference environment are studied,namely the Sparse Spectrum Fitting-Matrix Filtering(SPSF-MF)and the Sparse Spectrum Fitting-Matrix Filtering with Nulling(SPSF-MFN).Based on their principles,performances of the two algorithms under different interference intensity and off-grid situation are analyzed,computational complexity is also analyzed.Results show that SPSF-MF uses stopband attenuation to remove interferences,but when interferences are strong,interferences cannot be completely suppressed,residual interferences will affect the localization accuracy of expected sources.Besides,SPSF-MF only considers on-grid targets,it is impossible to estimate off-grid expected sources with high precision.Although SPSF-MFN can completely remove strong interferences,it cannot accurately estimate DOAs of off-grid expected source.In addition,the two algorithms are based on convex optimization with high computational complexity which limits the application of the algorithms.2.An Off-grid Sparse Bayesian DOA estimation method based on Robust Orthogonal Matrix Filter with Nulling(OGSBI-ROMFN)is proposed.The proposed algorithm firstly improves MFN,and adds orthogonal constraints column by column.By means of the idea of robust beamforming,the proposed algorithm changes the design of each discrete grid point in the spatial domain into the design of continuous domain including this grid point,so it has better filtering effect for off-grid targets in the spatial domain.Finally,the proposed algorithm combines OGSBI to reconstruct sparse signals and obtain the DOA estimates.Simulation results show that compared with the existing two algorithms,the proposed algorithm has higher DOA estimation accuracy for off-grid targets under various conditions.3.A DOA estimation method for joint optimization of matrix filter and sparse dictionary is proposed.When the sparse dictionary is fixed,the matrix filter is designed column by column based on the Minimum Variance Distortionless Response(MVDR)criterion,which can reduce the complexity of the matrix filter while completely removing strong interferences.When the matrix filter is fixed,Orthogonal Matching Pursuit(OMP)and Stochastic Gradient Descent based on variable step size with Momentum Gradient(SGDMG)are used to complete sparse reconstruction and sparse dictionary learning respectively,grid points in the passband are also updated to make the design of the matrix filter in next iteration more precise.The theoretical analysis of the computational complexity of each algorithm proves that the proposed method has the lowest computational complexity.The simulation results show that the DOA estimation accuracy of the proposed algorithm is similar to that of OGSBI-ROMFN,but it consumes minimal operation time under various conditions.