Blind identification and separation methods for Linear-Quadratic mixtures and/or linearly independent non-stationary signals

This paper concerns blind mixture identification (BMI) and blind source separation (BSS). We consider non-stationary stochastic sources, more specifically sources with slight time-domain sparsity. We first propose a correlation-based BMI/BSS method for Linear-Quadratic mixtures, called LQ-TEMPCORR. We also investigate the applicability of this type of method to possibly statistically dependent (e.g. correlated) but linearly independent signals. We thus extend the scope of our linear instantaneous method LI-TEMPCORR as a spin-off of this new investigation.