Approximate maximum likelihood source separation using the natural gradient
Seungjin Choi, Andrzej S Cichocki, Liqing Zhang, Шун-ичи Амари · 2002
This paper addresses a maximum likelihood approach to source separation in the case of overdetermined mixtures corrupted by additive white Gaussian noise. We present an objective function that is an approximate likelihood function based on the Laplace approximation. Then we derive a natural gradient adaptation algorithm which maximizes the corresponding approximate likelihood function. Useful behavior of the proposed method is verified by numerical experiments.