A re-weighted algorithm for designing data dependent sensing dictionary

Abstract

The modified orthogonal matching pursuit (OMP) algorithm based on sensing dictionary, shows significant improvement for the performance of sparse recovery, especially in the case of highly coherent dictionary. Assuming a signal to be decomposed, a good sensing dictionary should depend not only on the ordinary dictionary but also the observed data. In this paper, a re-weighted algorithm for designing data dependent sensing dictionary is proposed by introducing the effective posteriori knowledge obtained from the observed data. Simulation results are presented to demonstrate the superior performance of data dependent sensing dictionary designed by the proposed algorithm. Key words: Coherent dictionary, data dependent sensing dictionary, modified orthogonal matching pursuit, sparse recovery.