Blind Separation of Cyclo-stationary Signals using Generalized Eigenvalue Approach

This communication addresses the problem of the blind separation of cyclo-stationary source signals in the case of an instantaneous linear mixture. We show that the separation can be realized by the generalized eigenvectors that simultaneously diagonalize the cyclic correlation matrices of the observation signals. The mixture matrix is estimated by the eigenvectors of the pencil matrix constructed with two different cyclic correlation matrices of the observation signals. We also discuss the extension of some existing second-order blind source separation methods. The main advantages of the proposed method are that a pre-whitening stage is dropped, the cyclic frequencies are unknown and it is consistent in noisy case. Numerical simulation results are presented to illustrate the performance of the proposed method in digital communications context.