Deep Scattering Spectrum

TL;DR

The paper introduces a deep scattering spectrum that extends traditional audio representations with wavelet-based transforms, achieving state-of-the-art results in music genre and phoneme classification tasks.

Contribution

It proposes a novel scattering transform framework that captures translation and deformation invariances, improving audio classification performance.

Findings

Achieved state-of-the-art accuracy on GTZAN genre classification.

Attained top results on TIMIT phoneme classification.

Demonstrated robustness to time-warping and frequency transposition.

Abstract

A scattering transform defines a locally translation invariant representation which is stable to time-warping deformations. It extends MFCC representations by computing modulation spectrum coefficients of multiple orders, through cascades of wavelet convolutions and modulus operators. Second-order scattering coefficients characterize transient phenomena such as attacks and amplitude modulation. A frequency transposition invariant representation is obtained by applying a scattering transform along log-frequency. State-the-of-art classification results are obtained for musical genre and phone classification on GTZAN and TIMIT databases, respectively.