Recovery of sparse signals using parallel look ahead orthogonal matching pursuit algorithm

For compressed sensing signal recovery, the look ahead orthogonal matching pursuit (LAOMP) algorithm exhibits excellent recovery accuracy but with high computational complexity. In order to reduce the computational complexity without losing too much recovery accuracy, this paper thus proposes a Parallel-LAOMP algorithm, which uses parallel atoms selection strategy corresponding to four atoms selection criteria. The contributions of this atom selection method are twofold. First, the combination of these four atoms selection criteria improves the correct rate of atoms selection during the first few iterations. Second, these four criteria all adopt a parallel selection method to reduce the computational burden. Compared with the LAOMP algorithm, Monte Carlo simulation results demonstrate that in clean and noisy measurement cases for Gaussian sparse signals, Parallel-LAOMP significantly reduces the computational burden and exhibits higher recovery accuracy simultaneously.