A Fast Group Sparsity Based Phase Retrieval Algorithm for Non-Coherent DOA Estimation

摘要

A non-coherent DOA estimation algorithm for multiple snapshots is presented. The non-coherent DOA estimation problem is firstly formulated as a non-convex joint group sparsity based phase retrieval problem, and then replaced by its convex surrogate alternative through applying the majorization-minimization technique, and finally the proximal gradient method is employed to solve the surrogate problem. The proposed algorithm is referred to as fasT grOup sparsitY Based phAse Retrieval (ToyBar). Compared to the state of the art, the proposed one achieves a better performance with a lower computational complexity, as demonstrated by computer simulations.