Modified orthogonal matching pursuit for multiple measurement vector with joint sparsity in super-resolution compressed sensing

The mismatch model errors incur after discretization in many compressed sensing (CS) applications, e.g., target discrimination, direction-of-arrival (DOA) estimation because the true parameters do not lie on the discretized grids. Superresolution CS techniques aim to reduce these errors through the design of discrete Fourier transformation (DFT) sensing matrix with a finer grid system. However, without hardware updates, the discretization with a large refinement factor leads to high coherence of sensing matrix and consequently poor reconstruction performances. Thus, this paper uses Multiple Measurement Vector (MMV) technique for joint-sparse recovery of super-resolution CS-DFT problems and simultaneously proposes a new joint-sparse MMV reconstruction variant of Orthogonal Matching Pursuit (OMPMMV) to improve their performances. Furthermore, we also propose Cyclic OMPMMV with an additional atom updating procedure to correct the wrongly estimated non-zero indices. Finally, the achieved results will be demonstrated through numerical experiments.