A perceptual representation of audio for co-channel source separation

Abstract

Jntroduction Despite the many advances in signal processing mathematics in recent years, the greatest gains and breakthroughs are made by exploiting special properties of the particular problem at hand. Since very often the ultimate destination of processed sound is a human listener, the many complex interactions and constraints of the auditory system are available for exploitation. However, these constraints are so involved that we are only just beginning to understand the possibilities they offer.