Abstract:

Considering the disadvantages of ignoring signal’s structured sparsity and the high complexity in high iterative layers in multipath matching pursuit (MMP), the block pruning multipath matching pursuit (BPMMP) is proposed to reconstruct the block-sparse signal. In this algorithm, an atomic block serves as a node in the path expansion, and branch pruning operation is introduced after a certain number of iterations. Thus, BPMMP reduces the data processing cost greatly. Moreover, for multiple measurement vector (MMV) problem, BPMMP for MMV (BPMMPMMV) is proposed. It can achieve joint signal reconstruction for multiple sensors within a small range in the wireless sensor network. Experimental results show that BPMMP outperforms MMP on the reconstruction performance, and BPMMPMMV achieves higher joint reconstruction performance than block A* orthogonal matching pursuit for MMV, subspace matching pursuit for MMV and orthogonal matching pursuit for MMV.