Group delay based re-weighted sparse recovery algorithms for robust and high-resolution source separation in DOA framework

Abstract

Sparse Recovery (SR) algorithms have been used widely for direction-of-arrival (DOA) estimation in spatially contiguous plane wave for their robust performance. But these algorithms have proven to be computationally costly. With a few sensors and at low SNRs, the noise dominates the data singular vectors and the sparse estimation of contiguous sources is incorrect. The magnitude spectrum-based re-weighted sparse recovery (RWSR) algorithms improve the robustness by re-weighting the sparse estimates. However, their efficiency degrades with decreasing the number of sensors at low SNRs. Therefore, this paper exhibits the significance of the phase spectrum, in the form of group-delay, for sparse and robust source estimation using RWSR algorithms for spatially contiguous sources. Further, an optimal re-weighted methodology based on simultaneously minimizing average-root-mean-square-error and maximizing the probability of separation is also proposed. The simulation results are carried out for Gaussian noise to demonstrate the excellent performance of the proposed algorithms. Copyright © 2021 ISCA.