Performance of direction of arrival estimation based on support vector regression in impulsive noise environment

In this paper, the problem of estimating the direction of arrival (DOA) in impulsive noise environments is considered. One possible way to model the impulsive noise is to introduce symmetric ¿-stable (S¿S) distribution. Robust co variation based MUSIC (ROC-MUSIC) and Fractional lower moment based MUSIC (FLOM-MUSIC) can be used to estimate the DOA under these conditions. But those methods require eigenvalue decomposition and involve mass computational complexity. In order to reduce the computational burden, the problem of DOA of can be approximated as a nonlinear mapping by means of support vector regression. This paper discusses the application of SVR-based DOA in presence of impulsive noise. Moreover, performance of this new algorithm is analyzed by comparison. Computer simulation results verify the correctness and effectiveness of the proposed method.