Abstract

A block version of the orthogonal matching pursuit (OMP) algorithm, termed BOMP, can yield better reconstruction performance for block-sparse signals than conventional algorithms. However, the redundant dictionaries for block sparse signals are block coherent in the particle applications. In this paper, we consider a modified version of BOMP by introducing the concept of sensing dictionary. Exploiting the alternating projection (AP), we propose a method to design sensing dictionary for this modified BOMP. Simulation results show that the modified BOMP with sensing dictionary provides significant improvement for the recovery performance in the case of block coherent dictionary.

Key words: Block-sparsity, sensing dictionary, orthogonal matching pursuit, sparse reconstruction.