5.1. Introduction

Until recently, the study of auditory processes has been mainly focused on perceptual qualities such as the pitch, loudness or timbre of a sound produced by a single source. Experimentations in psychoacoustics have brought to the fore a relationship between the physical properties of a sound and the feelings which it conveys; the physiological mechanisms underlying this relationship have been sketched. Models of auditory processes have been designed, which are based on the acoustic wave or its spectrum.

Unfortunately, the audio sources around us are rarely active in isolation. We live in a cacophony of superimposed voices, sounds and noise, the resulting spectrum of which is completely different from that of a single source. Each of our ears receives sound waves originating from a multitude of sources. However, it is generally possible to focus one's attention on a specific source and perceive its loudness, its pitch, its timbre, and understand what is being said (when dealing with a speech source), in spite of the presence of competing sounds. Conventional models, which have been designed for isolated sounds, are not powerful enough to account for the perception of multiple sources.

In his day, Helmholtz wondered how we could perceive individual qualities of instruments when they were played together [HEL 77]. However, it is with the works of Bregman that auditory scene analysis (or ASA) became a research topic in itself [BRE 90]. For Bregman, the study of the emergence of subjective...