Blind Identification of Underdetermined Mixtures Using Second-Order Cyclostationary Statistics

Aiming at the problem of blind identification of underdetermined mixtures, we propose a method of blind identification based on second-order cyclostationary statistics. Estimate non zero cycle frequencies of original signals by cycle auto-correlation function of mixtures first and then stack the cycle correlation matrices corresponding to different cycle frequencies and time lags in a three order tensor, final achieve the estimation of mixing matrix by canonical decomposition. Simulation results indicate that the proposed algorithm estimates the mixing matrix with higher accuracy compared to the conventional algorithms.