Application of wavelet array denoising to ESPRIT algorithm

ESPRIT algorithm has been widely used in direction-of-arrival (DOA) estimation problem. It degrades greatly in the presence of spatially correlated noises and closely spaced DOA's, and/or at low SNR. Filters are often used to harness the performance degradation because they can recover the signal from its noisy observation. Though Wiener filter is the optimal filter, it requires the knowledge of second-order statistics of the signals and noises, which is difficult to obtain in non-wide-sense stationary signals and noises. Wavelet array denoising has been introduced to ESPRIT method in this paper to provide a possible filtering method in spatially correlated noises. The proposed approach denoises the snapshots of each sensor in parallel and applies conventional ESPRIT algorithm to the denoised data matrix. Simulation results show that by using the proposed approach, we can gain advantages of estimation root mean square error (RMSE) reduction and resolution enhancement.