Research On Robust DOA Estimation Algorithm Based On Sparse Bayesian Learning

Direction-of-arrival(DOA)estimation has been drawn a considerable attention and plays an important role in array signal processing.It is widely used in civil and military fields,including radar target location,direction finding,mobile communication and electronic warfare,etc.After decades of development,a large number of algorithms have proposed for DOA estimation based on different theory.For example,the subspace-based algorithms Multiple Signal Classification(MUSIC)and Estimation of Signal Parameters via Rotational Invariance Techniques(ESPRIT),and compressed sensing(CS)based algorithms L1-norm optimization algorithm and sparse Bayesian learning(SBL)algorithm,etc.Moreover,it is continually developing towards low cost,high resolution and low computational complexity now.Compare with the subspace-based DOA estimation methods,the DOA estimation method based on CS theory can effectively overcome the problem of unsatisfactory performance of subspace algorithm under the condition of low snapshot number and low signal-to-noise-ratio(SNR).However,the performance of traditional sparse representation(SR)based DOA estimation algorithm will suffer from serious degradation in the case of array mutual coupling,non-uniform noise and off-grid error.Based on the sparse Bayesian learning theory,our research mainly studies the robust DOA estimation in the case of off-grid error respectively coexists with non-uniform noise and array mutual coupling.The main contribution of this research including:Firstly,the basic principle of DOA estimation is explained,the specific array signal model of DOA estimation is given,and the Bayes' theorem and the CS theory are systematically introduced.Secondly,the off-grid DOA estimation under mutual coupling conditions is studied,and a root sparse Bayesian algorithm with non-uniform noise is proposed.The banded complex symmetric Toeplitz structure of the mutual coupling matrix is adopted to eliminate the negative influence of array mutual coupling on DOA estimation.Then,the spatial discrete grid points are updated by finding the roots of a polynomial,so as to achieve offgrid DOA estimation.the DOA estimation is realized by utilizing the root sparing Bayesian learning strategy.Then,for the off-grid DOA estimation under non-uniform noise conditions,an off-grid DOA estimation algorithm under non-uniform noise condition is proposed.The covariance matrix of non-uniform noise is reconstructed by a modified inverse iterative method,thereby eliminating the influence of non-uniform noise on DOA estimation performance.Finally,for the practical application situation,the off-grid DOA estimation method under non-uniform noise condition is further studied.A novel off-grid DOA estimation algorithm based on Multiple Input Multiple Output(MIMO)arrays with non-uniform noise is proposed,where the covariance matrix of non-uniform noise is reconstructed by Least Square(LS)strategy.Furthermore,based on this algorithm,an assistant vehicle localization algorithm based on three collaborative base stations is proposed.A large number of simulation experiments are carried out for different DOA estimation algorithms in this research,and the simulation results are effective to prove the effectiveness and robustness of the proposed algorithms.