Research On DOA Estimation Based On Grid-dismatch Compressive Sensing

Compressed Sensing Direction of Arrival(CS-DOA)is a new technology for Direction of Arrival estimation.It's based on the assumption that the distribution of direction of arrival is sparse,and then uses the signal reconstruction methods to estimate DOAs.However,the estimation results of previous methods are based on predefined grid,which means that their estimation accuracy and resolution are restricted by discretization of spatial domain.In fact,the real value cannot just be on the grid point which is also called problem of grid mismatch.The accuracy and resolution of estimation will fade if the grid is coarse.When dense grid is applied,the steering matrix in signal model will go against Restricted Isometry Property(RIP),which may lead the estimation method failed.This paper focuses on the influence of discretization on DOA estimation and the method to decrease this kind of influence.A research on theory of compressed sensing is carried.And then several classic CS-DOA estimation methods are introduced,including L1-SVD,FOCUSS,IAA and Bayesian compressed sensing(BCS)based DOA estimation method.Simulation results show advantages and drawbacks of these methods.Aiming to solve the problem of grid mismatch,two categories of methods,which need a predefined grid,are discussed.That is,grid refinement method and off-grid method.Grid refinement method is a natural way to refine some of coarse grid to become dense,leading accuracy improvement.Off-grid method is based on a novel model called off grid model and utilizes BCS to estimate DOAs.Simulation results validate these methods.Furthermore,a grid-less method is studied,i.e.,sparse and parametric algorithm(SPA).This method doesn't need a predefined grid or discretization of the spatial domain.This paper compares SPA with the previous methods,and shows its advantages.