English

Joint Time-Frequency Scattering for Audio Classification

Sound 2018-08-06 v1

Abstract

We introduce the joint time-frequency scattering transform, a time shift invariant descriptor of time-frequency structure for audio classification. It is obtained by applying a two-dimensional wavelet transform in time and log-frequency to a time-frequency wavelet scalogram. We show that this descriptor successfully characterizes complex time-frequency phenomena such as time-varying filters and frequency modulated excitations. State-of-the-art results are achieved for signal reconstruction and phone segment classification on the TIMIT dataset.

Keywords

Cite

@article{arxiv.1512.02125,
  title  = {Joint Time-Frequency Scattering for Audio Classification},
  author = {Joakim Andén and Vincent Lostanlen and Stéphane Mallat},
  journal= {arXiv preprint arXiv:1512.02125},
  year   = {2018}
}

Comments

6 pages, 2 figures in IEEE 25th International Workshop on Machine Learning for Signal Processing (MLSP), 2015. Sept. 17-20. Boston, USA

R2 v1 2026-06-22T12:03:26.215Z