Blind source separation based on the support recovery of pilot-signals

Blind source separation (BSS) has been widely discussed since it has many real applications. Recently, under the assumption that mixing matrix is orthogonal and source signals are sparse, Mishali et al. developed an amazing BSS method by using the support recovery of sources and the singular value decomposition (SVD). However, the performance of the algorithm is not as good as expected. In this paper, we present a novel BSS method that is performed by an identification of the mixing matrix by introducing the so-called pilot-signals. The pilot-signals are not required to be known, rather, they are required to have a known extent of sparsity. The method includes two phases, the mixing matrix estimation and the separation phases. The estimation phase is constructed with iterating of the three parts, support recovery, mixing matrix identification and pilot-signals recovery. The numerical experiments show that proposed method can efficiently converge and can recover the unknown source signals efficiently.