Blind separation of cyclostationary signals from instantaneous mixtures

This paper presents a new approach for blind separation of unknown cyclostationary signals from instantaneous mixtures. The proposed method can perfectly separate the mixed source signals so long as they have either different cyclic frequencies or clock phases. This is a weaker condition than those required by the algorithms. The separation criterion is to diagonalize a polynomial matrix whose coefficient matrices consist of the correlation and cyclic correlation matrices, at time delay τ=0, of multiple measurements.