Underdetermined Sparse Blind Source Separation by Clustering on Hyperplanes

In the underdetermined blind source separation and sparse component analysis, we get sensor signals X = AS, where x isin R^mxn, A isin R^mxn, S isin R^mxn, because the mixed matrix A and source signals S aren´t known and m < n , namely, the number of sensor signals less than that of source signals, but we can know source signals are sparse, so we use the information to recover source signals by estimating the mixed matrix. This paper gives an algorithm for estimating mixed matrix based on sparse sources information in underdetermined blind separation by clustering on hyperplanes´ normal lines, and the good performance is shown by the last example.