English title

In this paper we propose the spatial sparsity based WCSSDOA method for multi speakers' Direction of arrival estimation. In the proposed method the sparse modeling is done based on the sensor signals' correlation matrix, which leads to low computational complexity. In this method the SVD decomposition of the noise covariance matrix is proposed to reach the free noise sparse model, which leads to high accurate method in the low SNR with respect to other methods. Simulation results show that contrary to MUSIC, Minimum-norm, and SSDOA methods, the proposed method could detect the direction of two narrowband sources with  distance in -5dB SNR, six narrowband sources with  distance in -15dB SNR, and two speakers with  distance in 10dB SNR. Also, the computational complexity of the proposed method is reduced 45-85% with respect to SSDOA method.