A Novel Adaptive Algorithm for Blind Signal Separation of Linear Mixture

A novel adaptive algorithm for blind separation of the linearly mixed signals based on the relationship of the variability between the source and the mixed signals is presented, which utilizes the generalized eigen-subspace decomposition based on the recursive least square (RLS) parallel computation. The presented algorithm has less mean square error by avoiding the complexity of the estimation of the high-order statistics of the source signal, and it has faster speed of convergence. Simulation results illustrate the efficiency of the algorithm.