Optimizing self-organizing timbre maps

two approaches

Petri Toiviainen

pp. 335-350

in: Marc Leman (ed), Music, Gestalt, and computing, Berlin, Springer, 1997

Abstract

The effect of using different auditory images and distance metrics on the final configuration of a self-organized timbre map is examined by comparing distance matrices obtained from simulations with a similarity rating matrix, obtained using the same set of stimuli as in the simulations. Two approaches are described. In the static approach, each stimulus is represented as a single multi-component vector. Gradient images, which are intended to represent idealizations of physiological gradient maps in the auditory pathway, are constructed. The optimal auditory image and distance metric, with respect to the similarity rating data, are searched using the gradient method. In the dynamic approach, each input stimulus is represented by a set of spectral vectors.The response patterns are constructed through temporal integration. The onset portions of the stimuli are emphasized using a scaling procedure which operates qualitatively in the same way as lateral inhibition.