MMSE-based version of OMP for recovery of discrete-valued sparse signals

Abstract

An enhanced algorithm for the reconstruction of discrete-valued sparse vectors from an underdetermined system of linear equations under the presence of noise is presented. Knowledge from the field of compressed sensing and from digital communications is combined in optimising the estimation step in the orthogonal matching pursuit (OMP) algorithm according to the minimum mean-squared error (MMSE) criterion. Thereby, a gain in performance of more than 1 dB is achieved compared with the conventional OMP.