Expectation maximisation-based approach to recovering multiple sparse signals with common sparsity pattern
The problem of simultaneously recovering multiple sparse signals bearing a common sparsity pattern is addressed. Specifically, a common Gaussian prior to all the sparse signals under consideration is assigned. This can make that the signals share the same sparsity pattern. Then, an expectation maximisation (EM)-based approach to learn the priori parameters from measurements, thereby leading to the recovery of sparse signals is adopted. Simulations verify that the proposed EM approach outperforms the state-of-the-art counterparts.