Blind Source Separation with Compressively Sensed Based on SBL.

In this paper an efficient method is proposed for blind separating and reconstructing signals from compressively sensed linear mixtures. We assume that all sources share a common sparse representation basis and nonzero indices of these signals are different mostly. Sparse Bayesian learning (SBL) is developed to estimate the product of mixture matrix and signal. Subsequently, we use clustering method to estimate the mixture matrix and use subspace algorithm to estimate all source signals. Numerical results show efficiency of the proposed algorithm compared to previous method.