Sparse Signal Recovery via Multi-Residual Based Greedy Method

To recover a sparse signal from compressive measurements has drawn significant interest of researchers in signal processing and mathematics communities in the last few years. Greedy methods are one class of the most commonly used algorithms in solving the sparse signal recovery problem due to its iterative characteristics and hence of low computational complexity. This paper extends the single residual in conventional greedy methods to a definition of multiple residuals, and based on this so called multi-residual criterion a type of greedy method is proposed. Simulation results show the advantage of the proposed method in recovery probability over several classic greedy methods, such as orthogonal matching pursuit (OMP).