Direction Of Arrival Estimation Methods Based On Sparse Representation

Direction of arrival(DOA)estimation is an important research area of array signal processing(ASP),which means using the sensor array with a known shape to estimate the source direction.Currently,DOA estimation algorithms have been widely used in radar,communication,sonar,speech,and other fields.Considering the natural spatial sparsity of array signals,in recent years,many scholars try to use the sparse representation framework to express DOA estimation tasks,improve the algorithms by using the structural information,and reduce the requirements for signal-to-noise ratio and the number of snapshots.So far,many approaches based on sparse representation have been proposed,but there is still room for further improvement,especially the extended algorithms on observation with specific forms.In this paper,we design DOA estimation algorithms for observation with specific forms based on sparse representation.The main work is summarized as follows:(1)Wideband DOA estimation via variational Bayes expectation-maximization.In this work,we use wideband array observation data to estimate DOA based on the sparse Bayesian learning(SBL)framework.To use the wideband information reasonably and efficiently,we assume that the power of signal and noise in each sub-band is different,and introduce indicator variables acting on all sub-bands to represent the occupation of the candidate directions.Additionally,we introduce the offset parameters to dynamically update the searching grids and the extended array manifold to realize off-grid DOA estimation.The introduction of indicator variables and some model parameters makes reasonable use of the joint spatial sparsity of wideband signals and improves the performance of the algorithm.The final wideband spatial spectrum is expressed as the sum of narrowband source power weighted by indicator variables.Simulation results show that the proposed method has higher accuracy than the comparison approaches.And it is more suitable for observation with a low signal-to-noise ratio(SNR)and few snapshots.The computational efficiency of our method is higher than that of the similar approach.(2)One-bit DOA estimation via improved complex-valued binary iterative hard thresholding.In this work,we use the 1-bit quantized array observation data to estimate DOA based on the iterative hard thresholding(IHT)framework.In this approach,we first define the reconstruction error of the row sparse candidate source matrix and use the gradient descent method to update the candidate source matrix.Then,a refined hard threshold function is utilized to promote the row sparsity of the candidate source matrix.A stopping criterion is designed to judge the convergence.To get sharper spatial spectrum peaks and improve the accuracy of estimation,we introduce the basis updating strategy to dynamically update the searching grids and the expanded array manifold.In addition,a backtracking strategy is introduced to recover some rows of the candidate source matrix to promote convergence.Simulation results show that the accuracy of this algorithm is higher than that of the comparison algorithms.And it is more suitable for the observation with low SNR and few snapshots.The computational efficiency of our approach is higher than that of the similar algorithm.