Gridless Method For Direction Of Arrival Estimation Using Toeplitz Characteristics

Array-based direction of arrival(DOA)estimation is a very popular area of array signal processing.The traditional beamforming based spatial spectrum estimation method and the subspace based DOA estimationmethod need to rasterize the spatial domain search angle when performing spectral search.The sparse recovery class method appears later.Compared with the subspace-based method,the sparse method is applicable to a wide range of application scenarios,but its main disadvantages cannot be ignored.In order to achieve sparsity,the continuous angular space must be divided by a predetermined set of rasters.The grid partition is too dense.The computational complexity is too high or even satisfies the RIP observation conditions.The accuracy of the grid partitioning too sparse algorithm will be seriously reduced.However,the DOA of the signal source belongs to a continuous angular space rather than a discrete angular space.All methods based on raster partitioning inevitably introduce bias and greatly affect the estimated performance.In view of the above problems,this thesis studies the method of gridless DOA.The work is as follows.(1)This thesis analyzes the array signal model and the traditional subspace algorithm MUSIC method and the gridless ESPRIT method.This paper describes an improved raster-like DOA algorithm called the sparse iterative covariance estimation method.The principle of this method is gridless and is implemented using a rasterized iterative approach.Finally,a non-grid DOA estimation method called A Discretization-Free Sparse and Parametric Approach for Linear Array Signal Processing(SPA)is introduced.(2)This thesis proposes a T-matrix reconstruction-free grid-free DOA estimation method for uniform linear arrays and sparse linear arrays based on uniform linear array extraction.The method is divided into two large processing flows.Firstly,the data covariance matrix is received by the array,and a fast reconstruction method of the Topplitz chemical covariance matrix is proposed.Then the data covariance matrix is used to select the hermitian_toeplitz structure,the specific reconstruction process is transformed into a semi-definite programming problem and solved;secondly,a post-processing technique based on the van dermund matrix decomposition theorem is used to estimate the parameters of interest from the reconstructed covariance matrix without grid,including the source position.,source power and noise variance.The numerical simulation shows that the proposed method has the same performance as the SPA method and the time consumption is better than the latter method,and the improvement is 20%.The proposed method is based on convex optimization,does not require source number estimation,and is also applicable to any number of snapshots,and is robust to related sources.(3)In this thesis,the problem of amplitude and phase error is considered,and the proposed method is robust.Finally,the hardware processing architecture and hardware array signal processing board based on FPGA and DSP work together are designed.Through the actual radar signal test,the signal acquisition performance of the hardware architecture designed in this thesis is very good,and it fully meets the design index.