Jon, There are many different approaches to this. In a recently finished EU project, BLISS, one of the issues they aimed for was source separation in complex auditory mixtures. Please see http://www.lis.inpg.fr/pages_perso/bliss/ for further details. /Mikael