Semi-blind signal extraction from instantaneous mixtures by combining non-Gaussianity and cyclostationary property

Independent Component Analysis (ICA) is a technique which separates statistically independent component from a set of measured signals. However, in many applications where prefer to extract only one desired signal, it requires an extra post-selection method. This paper develops a sequential blind signal extraction algorithm that attempts to extract cyclostationary sources with unique cyclic frequencies from instantaneous mixtures. The approach incorporates cyclostationarity constraint into the ICA process, using Newton-like optimization method. Simulation results demonstrate the efficacy of the proposed algorithm.