Real-time speech separation by semi-supervised nonnegative matrix factorization

Cyril Joder; Felix Weninger; Florian Eyben; David Virette; Björn Schuller

doi:10.1007/978-3-642-28551-6_40

Real-time speech separation by semi-supervised nonnegative matrix factorization

Cyril Joder, Felix Weninger, Florian Eyben, David Virette, Björn Schuller

Chair of Health Informatics

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

57 Scopus citations

Abstract

In this paper, we present an on-line semi-supervised algorithm for real-time separation of speech and background noise. The proposed system is based on Nonnegative Matrix Factorization (NMF), where fixed speech bases are learned from training data whereas the noise components are estimated in real-time on the recent past. Experiments with spontaneous conversational speech and real-life non-stationary noise show that this system performs as well as a supervised NMF algorithm exploiting noise components learned from the same noise environment as the test sample. Furthermore, it outperforms a supervised system trained on different noise conditions.

Original language	English
Title of host publication	Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings
Pages	322-329
Number of pages	8
DOIs	https://doi.org/10.1007/978-3-642-28551-6_40
State	Published - 2012
Event	10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012 - Tel Aviv, Israel Duration: 12 Mar 2012 → 15 Mar 2012

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	7191 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012
Country/Territory	Israel
City	Tel Aviv
Period	12/03/12 → 15/03/12

Access to Document

10.1007/978-3-642-28551-6_40

Cite this

Joder, C., Weninger, F., Eyben, F., Virette, D., & Schuller, B. (2012). Real-time speech separation by semi-supervised nonnegative matrix factorization. In Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings (pp. 322-329). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7191 LNCS). https://doi.org/10.1007/978-3-642-28551-6_40

Joder, Cyril ; Weninger, Felix ; Eyben, Florian et al. / Real-time speech separation by semi-supervised nonnegative matrix factorization. Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings. 2012. pp. 322-329 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{20ed34c9c22449bb897ea89fca6c9004,

title = "Real-time speech separation by semi-supervised nonnegative matrix factorization",

abstract = "In this paper, we present an on-line semi-supervised algorithm for real-time separation of speech and background noise. The proposed system is based on Nonnegative Matrix Factorization (NMF), where fixed speech bases are learned from training data whereas the noise components are estimated in real-time on the recent past. Experiments with spontaneous conversational speech and real-life non-stationary noise show that this system performs as well as a supervised NMF algorithm exploiting noise components learned from the same noise environment as the test sample. Furthermore, it outperforms a supervised system trained on different noise conditions.",

author = "Cyril Joder and Felix Weninger and Florian Eyben and David Virette and Bj{\"o}rn Schuller",

year = "2012",

doi = "10.1007/978-3-642-28551-6_40",

language = "English",

isbn = "9783642285509",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

pages = "322--329",

booktitle = "Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings",

note = "10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012 ; Conference date: 12-03-2012 Through 15-03-2012",

}

Joder, C, Weninger, F, Eyben, F, Virette, D & Schuller, B 2012, Real-time speech separation by semi-supervised nonnegative matrix factorization. in Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 7191 LNCS, pp. 322-329, 10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012, Tel Aviv, Israel, 12/03/12. https://doi.org/10.1007/978-3-642-28551-6_40

Real-time speech separation by semi-supervised nonnegative matrix factorization. / Joder, Cyril; Weninger, Felix; Eyben, Florian et al.
Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings. 2012. p. 322-329 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7191 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Real-time speech separation by semi-supervised nonnegative matrix factorization

AU - Joder, Cyril

AU - Weninger, Felix

AU - Eyben, Florian

AU - Virette, David

AU - Schuller, Björn

PY - 2012

Y1 - 2012

N2 - In this paper, we present an on-line semi-supervised algorithm for real-time separation of speech and background noise. The proposed system is based on Nonnegative Matrix Factorization (NMF), where fixed speech bases are learned from training data whereas the noise components are estimated in real-time on the recent past. Experiments with spontaneous conversational speech and real-life non-stationary noise show that this system performs as well as a supervised NMF algorithm exploiting noise components learned from the same noise environment as the test sample. Furthermore, it outperforms a supervised system trained on different noise conditions.

AB - In this paper, we present an on-line semi-supervised algorithm for real-time separation of speech and background noise. The proposed system is based on Nonnegative Matrix Factorization (NMF), where fixed speech bases are learned from training data whereas the noise components are estimated in real-time on the recent past. Experiments with spontaneous conversational speech and real-life non-stationary noise show that this system performs as well as a supervised NMF algorithm exploiting noise components learned from the same noise environment as the test sample. Furthermore, it outperforms a supervised system trained on different noise conditions.

UR - http://www.scopus.com/inward/record.url?scp=84857354411&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-28551-6_40

DO - 10.1007/978-3-642-28551-6_40

M3 - Conference contribution

AN - SCOPUS:84857354411

SN - 9783642285509

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 322

EP - 329

BT - Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings

T2 - 10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012

Y2 - 12 March 2012 through 15 March 2012

ER -

Joder C, Weninger F, Eyben F, Virette D, Schuller B. Real-time speech separation by semi-supervised nonnegative matrix factorization. In Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings. 2012. p. 322-329. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-642-28551-6_40

Real-time speech separation by semi-supervised nonnegative matrix factorization

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this