Abstract

The Ear's Mind is principally based on the Copycat computer model (Mitchell 1993, Hofstadter 1995). The Ear's Mind is designed to model the unconscious, automatic auditory grouping pressures in humans. Such pressures, it seems, steer the perception of sound by cooperative and competitive interactions, resulting in the grouping of sound elements into context-sensible entities (see Bregman 1990, Rosenthal 1998, Ehret 1997). The auditory aspects of the proposed model are based on psychoacoustic experiments taken from the field of Auditory Scene Analysis (Bregman 1990). Like a community of social insects, where high-level phenomena emerge from local activities at lower-levels in a process of self-organisation, an interpretation of the auditory scene conforming human auditory perception is meant to emerge from the local activities of The Ear's Mind's agents. Although the current implementation is tailored for the auditory domain, the underlying theory of The Ear's Mind is meant as a general architecture for emergent sensory perception. Preliminary results support the theory and show both flexible context-sensitivity and the ability to match simple grouping behaviour. Further model expansion is planned.