Sparse Gabor multiplier estimation for identification of sound objects in texture sound

Monika Dörfler

Ewa Matusiak

pp. 443-462

in: Mitsuko Aramaki, Olivier Derrien, Richard Kronland-Martinet, Sølvi Ystad (eds), Sound, music, and motion, Berlin, Springer, 2014

Abstract

In this contribution we present a novel method for identifying novelty and, more specifically, sound objects within texture sounds. We introduce the notion of texture sound and sound object and explain how the properties of a sound that is known to be textural may be exploited in order to detect deviations which suggest the presence of novelty or distinct sound event, which may then be called sound object. The suggested approach is based on Gabor multipliers, which map the Gabor coefficients corresponding to certain time-segments of the signal to each other. We present the results of simulations based on both synthetic and real audio signals.