Publication result detail

Gabor frames and deep scattering networks in audio processing

BAMMER, R.; DÖRFLER, M.; HARÁR, P.

Original Title

Gabor frames and deep scattering networks in audio processing

English Title

Gabor frames and deep scattering networks in audio processing

Type

WoS Article

Original Abstract

This paper introduces Gabor scattering, a feature extractor based on Gabor frames and Mallat's scattering transform. By using a simple signal model for audio signals specific properties of Gabor scattering are studied. It is shown that for each layer, specific invariances to certain signal characteristics occur. Furthermore, deformation stability of the coefficient vector generated by the feature extractor is derived by using a decoupling technique which exploits the contractivity of general scattering networks. Deformations are introduced as changes in spectral shape and frequency modulation. The theoretical results are illustrated by numerical examples and experiments. Numerical evidence is given by evaluation on a synthetic and a "real" data set, that the invariances encoded by the Gabor scattering transform lead to higher performance in comparison with just using Gabor transform, especially when few training samples are available.

English abstract

This paper introduces Gabor scattering, a feature extractor based on Gabor frames and Mallat's scattering transform. By using a simple signal model for audio signals specific properties of Gabor scattering are studied. It is shown that for each layer, specific invariances to certain signal characteristics occur. Furthermore, deformation stability of the coefficient vector generated by the feature extractor is derived by using a decoupling technique which exploits the contractivity of general scattering networks. Deformations are introduced as changes in spectral shape and frequency modulation. The theoretical results are illustrated by numerical examples and experiments. Numerical evidence is given by evaluation on a synthetic and a "real" data set, that the invariances encoded by the Gabor scattering transform lead to higher performance in comparison with just using Gabor transform, especially when few training samples are available.

Keywords

machine learning; scattering transform; Gabor transform; deep learning; time-frequency analysis; CNN;

Key words in English

machine learning; scattering transform; Gabor transform; deep learning; time-frequency analysis; CNN;

Authors

BAMMER, R.; DÖRFLER, M.; HARÁR, P.

RIV year

2020

Released

26.09.2019

Publisher

MDPI

Location

Switzerland

ISBN

2075-1680

Periodical

Axioms

Volume

8

Number

4

State

Swiss Confederation

Pages from

1

Pages to

25

Pages count

25

URL

https://www.mdpi.com/2075-1680/8/4/106

Full text in the Digital Library

http://hdl.handle.net/11012/194791

BibTex

@article{BUT159057,
  author="Roswitha {Bammer} and Monika {Dörfler} and Pavol {Harár}",
  title="Gabor frames and deep scattering networks in audio processing",
  journal="Axioms",
  year="2019",
  volume="8",
  number="4",
  pages="1--25",
  doi="10.3390/axioms8040106",
  url="https://www.mdpi.com/2075-1680/8/4/106"
}

Documents

axioms-08-00106-v2