An Improved Spatial Smoothing Technique Based on Cross-covariance for Coherent Signals DOA Estimation

The spatial smoothing technique and its variants can resolve the coherent signals combined with the subspace-based methods and have a relatively low computational complexity. However, the existing spatial smoothing methods suffer from the low noise suppression ability, which further results in performance degradation of the DOA estimation for the low SNR regime. To overcome these problems, we propose a new spatial smoothing method to improve the DOA estimation performance of the coherent signals in the low SNR. Firstly, the whole array is split into several overlapped subarrays, and a full set of cross-covariance matrices are then computed via each individual subarray and its corresponding non-overlapping complementary subarray. This processing can enhance the noise suppression ability and further improve the SNR. Then, we employ the complete information of the cross-covariance matrices to achieve better estimates of the reconstructed covariance matrix, which will enhance the eigenvalue ratio of the signal-to-noise at low SNRs. Finally, the estimated DOAs can be obtained by combining with the subspace-based methods. Simulation results verify the superiority of the proposed method for the low SNRs.