On blind separation of convolutive mixtures of independent linear signals in unknown additive noise

Abstract

Blind separation of independent signals (sources) from their linear convolutive mixtures is considered. The various signals are assumed to be linear non-Gaussian but not necessarily i.i.d. First, an iterative, normalized higher order cumulant maximization-based approach is exploited using the third- and/or fourth-order normalized cumulants of the "beamformed" data. It provides a decomposition of the given data at each sensor into its independent signal components. In a second approach, higher order cumulant matching is used to consistently estimate the MIMO impulse response via nonlinear optimization. In a third approach, higher order cumulants are augmented with correlations. For blind signal separation, the estimated channel is used to decompose the received signal at each sensor into its independent signal components via a Wiener filter. Two illustrative simulation examples are presented.