Feature learning with deep scattering for urban sound analysis

Justin Salamon, Juan Pablo Bello

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this paper we evaluate the scattering transform as an alternative signal representation to the mel-spectrogram in the context of unsupervised feature learning for urban sound classification. We show that we can obtain comparable (or better) performance using the scattering transform whilst reducing both the amount of training data required for feature learning and the size of the learned codebook by an order of magnitude. In both cases the improvement is attributed to the local phase invariance of the representation. We also observe improved classification of sources in the background of the auditory scene, a result that provides further support for the importance of temporal modulation in sound segregation.

Original languageEnglish (US)
Title of host publication2015 23rd European Signal Processing Conference, EUSIPCO 2015
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages724-728
Number of pages5
ISBN (Electronic)9780992862633
DOIs
StatePublished - Dec 22 2015
Event23rd European Signal Processing Conference, EUSIPCO 2015 - Nice, France
Duration: Aug 31 2015Sep 4 2015

Other

Other23rd European Signal Processing Conference, EUSIPCO 2015
CountryFrance
CityNice
Period8/31/159/4/15

Fingerprint

Acoustic waves
Scattering
Invariance
Modulation

Keywords

  • acoustic event classification
  • machine learning
  • scattering transform
  • Unsupervised learning
  • urban

ASJC Scopus subject areas

  • Media Technology
  • Computer Vision and Pattern Recognition
  • Signal Processing

Cite this

Salamon, J., & Bello, J. P. (2015). Feature learning with deep scattering for urban sound analysis. In 2015 23rd European Signal Processing Conference, EUSIPCO 2015 (pp. 724-728). [7362478] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/EUSIPCO.2015.7362478

Feature learning with deep scattering for urban sound analysis. / Salamon, Justin; Bello, Juan Pablo.

2015 23rd European Signal Processing Conference, EUSIPCO 2015. Institute of Electrical and Electronics Engineers Inc., 2015. p. 724-728 7362478.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Salamon, J & Bello, JP 2015, Feature learning with deep scattering for urban sound analysis. in 2015 23rd European Signal Processing Conference, EUSIPCO 2015., 7362478, Institute of Electrical and Electronics Engineers Inc., pp. 724-728, 23rd European Signal Processing Conference, EUSIPCO 2015, Nice, France, 8/31/15. https://doi.org/10.1109/EUSIPCO.2015.7362478
Salamon J, Bello JP. Feature learning with deep scattering for urban sound analysis. In 2015 23rd European Signal Processing Conference, EUSIPCO 2015. Institute of Electrical and Electronics Engineers Inc. 2015. p. 724-728. 7362478 https://doi.org/10.1109/EUSIPCO.2015.7362478
Salamon, Justin ; Bello, Juan Pablo. / Feature learning with deep scattering for urban sound analysis. 2015 23rd European Signal Processing Conference, EUSIPCO 2015. Institute of Electrical and Electronics Engineers Inc., 2015. pp. 724-728
@inproceedings{1f84f282dc6e4905aa5e3293f814cca1,
title = "Feature learning with deep scattering for urban sound analysis",
abstract = "In this paper we evaluate the scattering transform as an alternative signal representation to the mel-spectrogram in the context of unsupervised feature learning for urban sound classification. We show that we can obtain comparable (or better) performance using the scattering transform whilst reducing both the amount of training data required for feature learning and the size of the learned codebook by an order of magnitude. In both cases the improvement is attributed to the local phase invariance of the representation. We also observe improved classification of sources in the background of the auditory scene, a result that provides further support for the importance of temporal modulation in sound segregation.",
keywords = "acoustic event classification, machine learning, scattering transform, Unsupervised learning, urban",
author = "Justin Salamon and Bello, {Juan Pablo}",
year = "2015",
month = "12",
day = "22",
doi = "10.1109/EUSIPCO.2015.7362478",
language = "English (US)",
pages = "724--728",
booktitle = "2015 23rd European Signal Processing Conference, EUSIPCO 2015",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
address = "United States",

}

TY - GEN

T1 - Feature learning with deep scattering for urban sound analysis

AU - Salamon, Justin

AU - Bello, Juan Pablo

PY - 2015/12/22

Y1 - 2015/12/22

N2 - In this paper we evaluate the scattering transform as an alternative signal representation to the mel-spectrogram in the context of unsupervised feature learning for urban sound classification. We show that we can obtain comparable (or better) performance using the scattering transform whilst reducing both the amount of training data required for feature learning and the size of the learned codebook by an order of magnitude. In both cases the improvement is attributed to the local phase invariance of the representation. We also observe improved classification of sources in the background of the auditory scene, a result that provides further support for the importance of temporal modulation in sound segregation.

AB - In this paper we evaluate the scattering transform as an alternative signal representation to the mel-spectrogram in the context of unsupervised feature learning for urban sound classification. We show that we can obtain comparable (or better) performance using the scattering transform whilst reducing both the amount of training data required for feature learning and the size of the learned codebook by an order of magnitude. In both cases the improvement is attributed to the local phase invariance of the representation. We also observe improved classification of sources in the background of the auditory scene, a result that provides further support for the importance of temporal modulation in sound segregation.

KW - acoustic event classification

KW - machine learning

KW - scattering transform

KW - Unsupervised learning

KW - urban

UR - http://www.scopus.com/inward/record.url?scp=84963983684&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84963983684&partnerID=8YFLogxK

U2 - 10.1109/EUSIPCO.2015.7362478

DO - 10.1109/EUSIPCO.2015.7362478

M3 - Conference contribution

SP - 724

EP - 728

BT - 2015 23rd European Signal Processing Conference, EUSIPCO 2015

PB - Institute of Electrical and Electronics Engineers Inc.

ER -