A GMM sound source model for blind speech separation in under-determined conditions

Yasuharu Hirasawa; Naoki Yasuraoka; Toru Takahashi; Tetsuya Ogata; Hiroshi G. Okuno

doi:10.1007/978-3-642-28551-6_55

A GMM sound source model for blind speech separation in under-determined conditions

Yasuharu Hirasawa^*, Naoki Yasuraoka, Toru Takahashi, Tetsuya Ogata, Hiroshi G. Okuno

^*Corresponding author for this work

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

2 Citations (Scopus)

Abstract

This paper focuses on blind speech separation in under-determined conditions, that is, in the case when there are more sound sources than microphones. We introduce a sound source model based on the Gaussian mixture model (GMM) to represent a speech signal in the time-frequency domain, and derive rules for updating the model parameters using the auxiliary function method. Our GMM sound source model consists of two kinds of Gaussians: sharp ones representing harmonic parts and smooth ones representing nonharmonic parts. Experimental results reveal that our method outperforms the method based on non-negative matrix factorization (NMF) by 0.7dB in the signal-to-distortion ratio (SDR), and by 1.7dB in the signal-to-interference ratio (SIR). This means that our method effectively removes interference coming from other talkers.

Original language	English
Title of host publication	Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings
Pages	446-453
Number of pages	8
DOIs	https://doi.org/10.1007/978-3-642-28551-6_55
Publication status	Published - 2012
Externally published	Yes
Event	10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012 - Tel Aviv, Israel Duration: 2012 Mar 12 → 2012 Mar 15

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	12/3/12 → 12/3/15

Keywords

Auxiliary function method
Blind speech separation
GMM sound source model
Under-determined condition

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-642-28551-6_55

Cite this

Hirasawa, Y., Yasuraoka, N., Takahashi, T., Ogata, T., & Okuno, H. G. (2012). A GMM sound source model for blind speech separation in under-determined conditions. In Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings (pp. 446-453). (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_55

A GMM sound source model for blind speech separation in under-determined conditions. / Hirasawa, Yasuharu; Yasuraoka, Naoki; Takahashi, Toru et al.
Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings. 2012. p. 446-453 (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

Hirasawa, Y, Yasuraoka, N, Takahashi, T, Ogata, T & Okuno, HG 2012, A GMM sound source model for blind speech separation in under-determined conditions. 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. 446-453, 10th International Conference on Latent Variable Analysis and Signal Separation, LVA/ICA 2012, Tel Aviv, Israel, 12/3/12. https://doi.org/10.1007/978-3-642-28551-6_55

Hirasawa Y, Yasuraoka N, Takahashi T, Ogata T, Okuno HG. A GMM sound source model for blind speech separation in under-determined conditions. In Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings. 2012. p. 446-453. (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_55

Hirasawa, Yasuharu ; Yasuraoka, Naoki ; Takahashi, Toru et al. / A GMM sound source model for blind speech separation in under-determined conditions. Latent Variable Analysis and Signal Separation - 10th International Conference, LVA/ICA 2012, Proceedings. 2012. pp. 446-453 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{7310fe4aba22462c86b794c5a752aefc,

title = "A GMM sound source model for blind speech separation in under-determined conditions",

abstract = "This paper focuses on blind speech separation in under-determined conditions, that is, in the case when there are more sound sources than microphones. We introduce a sound source model based on the Gaussian mixture model (GMM) to represent a speech signal in the time-frequency domain, and derive rules for updating the model parameters using the auxiliary function method. Our GMM sound source model consists of two kinds of Gaussians: sharp ones representing harmonic parts and smooth ones representing nonharmonic parts. Experimental results reveal that our method outperforms the method based on non-negative matrix factorization (NMF) by 0.7dB in the signal-to-distortion ratio (SDR), and by 1.7dB in the signal-to-interference ratio (SIR). This means that our method effectively removes interference coming from other talkers.",

keywords = "Auxiliary function method, Blind speech separation, GMM sound source model, Under-determined condition",

author = "Yasuharu Hirasawa and Naoki Yasuraoka and Toru Takahashi and Tetsuya Ogata and Okuno, {Hiroshi G.}",

year = "2012",

doi = "10.1007/978-3-642-28551-6_55",

language = "English",

isbn = "9783642285509",

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

pages = "446--453",

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",

}

TY - GEN

T1 - A GMM sound source model for blind speech separation in under-determined conditions

AU - Hirasawa, Yasuharu

AU - Yasuraoka, Naoki

AU - Takahashi, Toru

AU - Ogata, Tetsuya

AU - Okuno, Hiroshi G.

PY - 2012

Y1 - 2012

N2 - This paper focuses on blind speech separation in under-determined conditions, that is, in the case when there are more sound sources than microphones. We introduce a sound source model based on the Gaussian mixture model (GMM) to represent a speech signal in the time-frequency domain, and derive rules for updating the model parameters using the auxiliary function method. Our GMM sound source model consists of two kinds of Gaussians: sharp ones representing harmonic parts and smooth ones representing nonharmonic parts. Experimental results reveal that our method outperforms the method based on non-negative matrix factorization (NMF) by 0.7dB in the signal-to-distortion ratio (SDR), and by 1.7dB in the signal-to-interference ratio (SIR). This means that our method effectively removes interference coming from other talkers.

AB - This paper focuses on blind speech separation in under-determined conditions, that is, in the case when there are more sound sources than microphones. We introduce a sound source model based on the Gaussian mixture model (GMM) to represent a speech signal in the time-frequency domain, and derive rules for updating the model parameters using the auxiliary function method. Our GMM sound source model consists of two kinds of Gaussians: sharp ones representing harmonic parts and smooth ones representing nonharmonic parts. Experimental results reveal that our method outperforms the method based on non-negative matrix factorization (NMF) by 0.7dB in the signal-to-distortion ratio (SDR), and by 1.7dB in the signal-to-interference ratio (SIR). This means that our method effectively removes interference coming from other talkers.

KW - Auxiliary function method

KW - Blind speech separation

KW - GMM sound source model

KW - Under-determined condition

UR - http://www.scopus.com/inward/record.url?scp=84863115686&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84863115686&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-28551-6_55

DO - 10.1007/978-3-642-28551-6_55

M3 - Conference contribution

AN - SCOPUS:84863115686

SN - 9783642285509

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

SP - 446

EP - 453

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 -

A GMM sound source model for blind speech separation in under-determined conditions

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this