Automatic audio tagging using covariate shift adaptation

Gordon Wichern, Makoto Yamada, Harvey Thornburg, Masashi Sugiyama, Andreas Spanias

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

2 Citations (Scopus)

Abstract

Automatically annotating or tagging unlabeled audio files has several applications, such as database organization and recommender systems. We are interested in the case where the system is trained using clean high-quality audio files, butmost of the files that need to be automatically tagged during the test phase are heavily compressed and noisy, for instance if they were captured on a mobile device. In this situation we assume the audio files follow a covariate shift model in the acoustic feature space, i.e., the feature distributions are different in the training and test phases, but the conditional distribution of labels given features remains unchanged. Our method uses a specially designed audio similarity measure as input to a set of weighted logistic regressors, which attempt to alleviate the influence of covariate shift. Results on a freely available database of sound files contributed and labeled by non-expert users, demonstrate effective automatic tagging performance.

Original languageEnglish (US)
Title of host publicationICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Pages253-256
Number of pages4
DOIs
StatePublished - 2010
Event2010 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2010 - Dallas, TX, United States
Duration: Mar 14 2010Mar 19 2010

Other

Other2010 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2010
CountryUnited States
CityDallas, TX
Period3/14/103/19/10

Fingerprint

Recommender systems
Mobile devices
Logistics
Labels
Acoustics
Acoustic waves

Keywords

  • Acoustic signal analysis
  • Database query processing
  • Importance
  • KLIEP

ASJC Scopus subject areas

  • Signal Processing
  • Software
  • Electrical and Electronic Engineering

Cite this

Wichern, G., Yamada, M., Thornburg, H., Sugiyama, M., & Spanias, A. (2010). Automatic audio tagging using covariate shift adaptation. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings (pp. 253-256). [5495973] https://doi.org/10.1109/ICASSP.2010.5495973

Automatic audio tagging using covariate shift adaptation. / Wichern, Gordon; Yamada, Makoto; Thornburg, Harvey; Sugiyama, Masashi; Spanias, Andreas.

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. 2010. p. 253-256 5495973.

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

Wichern, G, Yamada, M, Thornburg, H, Sugiyama, M & Spanias, A 2010, Automatic audio tagging using covariate shift adaptation. in ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings., 5495973, pp. 253-256, 2010 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2010, Dallas, TX, United States, 3/14/10. https://doi.org/10.1109/ICASSP.2010.5495973
Wichern G, Yamada M, Thornburg H, Sugiyama M, Spanias A. Automatic audio tagging using covariate shift adaptation. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. 2010. p. 253-256. 5495973 https://doi.org/10.1109/ICASSP.2010.5495973
Wichern, Gordon ; Yamada, Makoto ; Thornburg, Harvey ; Sugiyama, Masashi ; Spanias, Andreas. / Automatic audio tagging using covariate shift adaptation. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. 2010. pp. 253-256
@inproceedings{9cdee75ea4c9406898c5906497b31275,
title = "Automatic audio tagging using covariate shift adaptation",
abstract = "Automatically annotating or tagging unlabeled audio files has several applications, such as database organization and recommender systems. We are interested in the case where the system is trained using clean high-quality audio files, butmost of the files that need to be automatically tagged during the test phase are heavily compressed and noisy, for instance if they were captured on a mobile device. In this situation we assume the audio files follow a covariate shift model in the acoustic feature space, i.e., the feature distributions are different in the training and test phases, but the conditional distribution of labels given features remains unchanged. Our method uses a specially designed audio similarity measure as input to a set of weighted logistic regressors, which attempt to alleviate the influence of covariate shift. Results on a freely available database of sound files contributed and labeled by non-expert users, demonstrate effective automatic tagging performance.",
keywords = "Acoustic signal analysis, Database query processing, Importance, KLIEP",
author = "Gordon Wichern and Makoto Yamada and Harvey Thornburg and Masashi Sugiyama and Andreas Spanias",
year = "2010",
doi = "10.1109/ICASSP.2010.5495973",
language = "English (US)",
isbn = "9781424442966",
pages = "253--256",
booktitle = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

}

TY - GEN

T1 - Automatic audio tagging using covariate shift adaptation

AU - Wichern, Gordon

AU - Yamada, Makoto

AU - Thornburg, Harvey

AU - Sugiyama, Masashi

AU - Spanias, Andreas

PY - 2010

Y1 - 2010

N2 - Automatically annotating or tagging unlabeled audio files has several applications, such as database organization and recommender systems. We are interested in the case where the system is trained using clean high-quality audio files, butmost of the files that need to be automatically tagged during the test phase are heavily compressed and noisy, for instance if they were captured on a mobile device. In this situation we assume the audio files follow a covariate shift model in the acoustic feature space, i.e., the feature distributions are different in the training and test phases, but the conditional distribution of labels given features remains unchanged. Our method uses a specially designed audio similarity measure as input to a set of weighted logistic regressors, which attempt to alleviate the influence of covariate shift. Results on a freely available database of sound files contributed and labeled by non-expert users, demonstrate effective automatic tagging performance.

AB - Automatically annotating or tagging unlabeled audio files has several applications, such as database organization and recommender systems. We are interested in the case where the system is trained using clean high-quality audio files, butmost of the files that need to be automatically tagged during the test phase are heavily compressed and noisy, for instance if they were captured on a mobile device. In this situation we assume the audio files follow a covariate shift model in the acoustic feature space, i.e., the feature distributions are different in the training and test phases, but the conditional distribution of labels given features remains unchanged. Our method uses a specially designed audio similarity measure as input to a set of weighted logistic regressors, which attempt to alleviate the influence of covariate shift. Results on a freely available database of sound files contributed and labeled by non-expert users, demonstrate effective automatic tagging performance.

KW - Acoustic signal analysis

KW - Database query processing

KW - Importance

KW - KLIEP

UR - http://www.scopus.com/inward/record.url?scp=78049412206&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78049412206&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2010.5495973

DO - 10.1109/ICASSP.2010.5495973

M3 - Conference contribution

AN - SCOPUS:78049412206

SN - 9781424442966

SP - 253

EP - 256

BT - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

ER -