Adaptive algorithm for the blind separation of sources with finite support

A new blind separation algorithm which exploits only the finite support property and the independence of the sources is proposed. The approach is based on the minimization of magnitude peaks of some affine mapping outputs. It is proven that the perfect separators are the only global minima of the corresponding cost function. The proposed method leads to the construction of a simple adaptive algorithm for the separation of sources. A simulation example is provided to illustrate the algorithm's performance.