An Adaptive Threshold Algorithm for Joint Sparse Recovery

Abstract

This paper presents a new greedy algorithm for joint sparse recovery, called adaptive threshold simultaneous orthogonal matching pursuit (AT-SOMP). In this algorithm, an adaptive threshold is designed based on subspace decomposition to stop the iterative process, during which the support of target signal is identified iteratively just like the simultaneous orthogonal matching pursuit (S-OMP) algorithm. As the adaptive threshold can change with the noise power adaptively, the proposed algorithm is applied to the situation that both the sparsity and SNR of the target signal are unknown. Experiments validate that in this situation the proposed algorithm has a better recovery performance than those methods with a fixed threshold. Keywords-compressed sensing; joint sparse recovery; adaptive threshold; simultaneous orthogonal matching pursuit