DOA Estimation Based On Sparse Signal Representation

Direction of arrival estimation as one of the main branches of array signal processing is widely used in many fields such as radar,communication and sonar,ect.Among the traditional DOA estimaiton methods,a class of most representative one is subspace-based method.However,a bit amount of snapshots and high signal-to-noise ratio are needed to ensure the accuracy of the estimation.Besides,the exactly prior information of the number of the signals is also required to construct the noise or signal subspace.The spatial sparsity of the sources makes it possible that DOA estimation can be applied in the frame of sparse signal representation.DOA estimation based on sparse signal representation exhibits many advantages over the classical subspace methods.Such as,it does not require the number of the sourses in advance and has better performance when the number of the snapshots is small and the signal-to-noise ratio is low,what is more,it can deal with coherent signals effectively.Therefore,the research of DOA estimation based on sparse signal representation is of great signification.The main work of this paper is divided into the following sections:1.The latest development about DOA estimation is summarized and the fundamental theories of array signal processing are presented.The models of narrowband and wideband of array signal are set up.Some classical algorithms of DOA estimation are introduced and their performances are analyzed.2.The basic theories of sparse signal representation are described.Three classical sparse reconstruction algorithms--Matching Pursuit(MP),Basis Pursuit(BP)and FOCUSS are introduced.The rationality of sparse signal representation in DOA estimation is analysised.3.The greedy block coordinate descent(GBCD)algorithm is a complex optimization problem,it suffers from a highly computational complexity and can’t deal with the coherent signal effectively.A new algorithm called fast weighted greedy block coordinate descent(FWGBCD)is proposed.By unitary transformation,we convert the complex optimization into a real one,thereby reducing the computational complexity without sacrificing its accuracy.The forward and backward spatial smoothing method has been used during the unitary transformation,therefor,the capacity of FWGBCD algorithm for two coherent signals processing has been further improved.4.The multiple dictionary joint sparse representation model for wideband signal processing is set up.A new algorithm-multiple-dictionary fast greedy block coordinate descent(MD-FGBCD)is proposed,which makes up for the shortcomings of l1-SVD in wideband applications.Firstly,the unitary transformation is used to convert the complex covariance matrix to the real one.Then,the multiple dictionaries joint sparse representation model is constructed.Finally,the idea of GBCD algorithm is used to reconstruct the sparse signal.Get the DOA information of the targets from the peaks of the sparse signal.Simulation results indicate the feasibility of the new algorithm.It has higher accuracy than l1-SVD and RSS,especially when the signal-to-noise is lower.