Non-stationary ICA
Richard Everson · 2004
Independent components analysis usually assumes that sources are mixed in constant proportions to form observations. However, this assumption is violated if the mixing proportions vary, as may occur in EEG measurements electrodes dry out or in audio mixing if sources move relative to the recording microphones. In this talk blind source separation with non-stationary mixing, but stationary sources is considered. The linear mixing of the independent sources is modelled as evolving according to a first order Markov process, and a method for tracking the mixing and simultaneously inferring the sources is presented. Observational noise is included in the model. The technique is illustrated with numerical examples. (15 pages)