Blind source separation of PSK and amplitude modulated signals

The paper proposes a new technique for blindly separating mixed independent PSK or QAM signals. Unlike in speech and image processing where the probability density function (PDF) of source signals is Gaussian, the PDF of PSK and QAM signals is non-Gaussian. Thus, direct application of conventional methods for blind source separation (BSS) of speech or image signals to the blind separation of PSK or QAM signals is not possible. It is therefore necessary to develop a function that can be integrated and applied to BSS for PSK and QAM signals. We propose a method which uses an approximate PDF for PSK and QAM signals. By simulation, we verified that this method effectively separates PSK or QAM signals that have been mixed in a Rayleigh flat-fading channel.