PROJECT TITLE :

Improving M-SBL for Joint Sparse Recovery Using a Subspace Penalty

ABSTRACT:

A multiple measurement vector drawback (MMV) may be a generalization of the compressed sensing drawback that addresses the recovery of a set of jointly sparse signal vectors. One amongst the important contributions of this paper is to show that the seemingly least connected state-of-the-art MMV joint sparse recovery algorithms—the M-SBL (multiple sparse Bayesian learning) and subspace-based mostly hybrid greedy algorithms—have a terribly vital link. A lot of specifically, we tend to show that replacing the $logdet(,cdot,)$ term in the M-SBL by a rank surrogate that exploits the spark reduction property discovered within the subspace-based joint sparse recovery algorithms provides significant improvements. In specific, if we use the Schatten-$p$ quasi-norm as the corresponding rank surrogate, the world minimizer of the price operate in the proposed algorithm becomes a twin of the true resolution as $p rightarrow zero$. Furthermore, underneath regularity conditions, we show that convergence to a native minimizer is guaranteed using an alternating minimization algorithm that has closed kind expressions for each of the minimization steps, which are convex. Numerical simulations underneath a variety of eventualities in terms of SNR and also the condition number of the signal amplitude matrix show that the proposed algorithm consistently outperformed the M-SBL and other state-of-the art algorithms.