Faster OMP computations by sensing matrix column reduction

Compressed sensing is an emerging technique that allows to reconstruct sparse signals sampled at sub-Nyquist rates. However, it requires high computational effort to reconstruct the compressively sampled signal, which makes real-time application of it very hard. We therefore, present a novel, generic method that decreases the computational complexity of Orthogonal Matehing Pursuit (OMP) like reconstruction algorithms that exploit the correlation of columns of a dictionary (sensing matrix). The proposed method reduces the column number of the dictionary in a systematic manner to speed up the correlation calculations. Simulation results show that in sparse scenarios, reconstruction speed increases significantiy with a negligible decrease in the reconstruction accuracy.