Transform\'ee en scattering sur la spirale temps-chroma-octave

TL;DR

This paper presents a novel scattering representation for sound analysis that is translation-invariant, deformation-stable, and captures harmonic structures, using a wavelet transform along a harmonic spiral.

Contribution

It introduces a new scattering method based on a wavelet transform along a harmonic spiral, enhancing sound analysis and classification capabilities.

Findings

Effective in analyzing source-filter model deformations

Captures harmonic structures robustly

Stable to temporal and frequency deformations

Abstract

We introduce a scattering representation for the analysis and classification of sounds. It is locally translation-invariant, stable to deformations in time and frequency, and has the ability to capture harmonic structures. The scattering representation can be interpreted as a convolutional neural network which cascades a wavelet transform in time and along a harmonic spiral. We study its application for the analysis of the deformations of the source-filter model.