DOA Estimation Of Coherent Signal Based On Sparse Bayesian Learning

The demand scenarios of direction-of-arrival(DOA)estimation is becoming more and more complicated,such as low SNR,coherent sources,multipath scattering phenomenon and so on.Traditional DOA estimation methods,such as subspace algorithms,have been difficult to achieve highly accurate DOA estimation.In view of the advantages of existing sparse reconstruction algorithms based on spatial sparsity,these sparse reconstruction algorithms are constrained by some factors,and there are still some shortcomings.Based on the spatial sparsity of the incident signal,this thesis studies the problem of fast and accurate DOA estimation by sparse Bayesian framework.The main research contents and innovations are as follows:1.The point source sparse representation model of the coherent sources and the traditional coherent distribution source model are studied.The intrinsic relation between sparse reconstruction based on 0? norm and maximum posteriori estimation is studied.The sparse Bayesian estimation is analyzed deeply,and its theoretical framework is given,and the relevant theory is given a detailed proof.We analyze the some properties of sparse Bayesian reconstruction algorithm in array signal processing.2.Based on the sparse Bayesian framework,a new method for the DOA estimation in presence of coherent sources is proposed.The difference technique is used to enhance the input SNR and a new spatial sampling method is used to form a complete dictionary.The DOA of the coherent sources are estimated via SBL with improved dictionary matrix and an effective DOA search method.The computational complexity of the proposed method is given and compared with the L1-SVD method and so on.The effectiveness of the proposed algorithm is verified by numerical simulation experiments.3.An efficient sparsity-inducing method for DOA estimation is proposed to solve the challenging problem of computation cost and resolution in the beam space.The array output is mapped to the beam space,and the covariance is expressed by the sparse representation of the over-complete dictionary at a larger angular interval.In doing so,the sparse Bayesian learning(SBL)technique is applied to enforce sparsity at the true source locations and the coarse sources locations are obtained.Then the refined method is used to get the high-resolution DOA estimation based on the coarse estimation.The computational complexity of the proposed method is given and compared with the L1-SRACV method and so on.The effectiveness of the proposed algorithm is verified by numerical simulation experiments.4.An efficient method for estimating the parameters of one-dimension coherent distribution sources based on sparse Bayesian is proposed.The sparse representation model of coherent distribution is studied,and the large sampling interval is used to sample the spatial and angular extension domains to form a generalized array manifold.On this basis,the method of estimating the parameters of coherent distribution sources by sparse Bayesian framework is proposed,and the angular parameter estimation of the distribution source is obtained by rough estimation and fine estimation.The computational complexity of the proposed method is given and compared with the DSPE method and so on.The effectiveness of the proposed algorithm is verified by numerical simulation experiments.