A new mixing matrix identification algorithm for underdetermined blind source separation

In this paper we focus on the mixing matrix identification problem for underdetermined blind source separation. Based on the two-stage approach in sparse component analysis, we proposal a new algorithm that integrate with other blind signal processing methods like independent component analysis and model order selection. Compared with the DUET, the TIFROM and standard clustering methods, this algorithm can work adaptively in noisy environment and the required sparseness of sources can be considerably relaxed. Simulation results are presented.