Joint Separation and Dereverberation of Reverberant Mixtures with Determined Multichannel Non-Negative Matrix Factorization

Hideaki Kagami, Hirokazu Kameoka, Masahiro Yukawa

研究成果: Conference contribution

4 引用 (Scopus)

抄録

This paper proposes an extension of multichannel non-negative matrix factorization (MNMF) that simultaneously solves source separation and dereverberation. While MNMF was originally formulated under an underdetermined problem setting where sources can outnumber microphones, a determined counterpart of MNMF, which we call the determined MNMF (DMNMF), has recently been proposed with notable success. This approach is particularly notable in that the optimization process can be more than 30 times faster than the underdetermined version owing to the fact that it involves no matrix inversion computations. One drawback as regards all methods based on instantaneous mixture models, including MNMF, is that they are weak against long reverberation. To overcome this drawback, this paper proposes an extension of DMNMF using a frequency-domain convolutive mixture model. The optimization process of the proposed method consists of iteratively updating (i) the spectral parameters of each source using the majorization-minimization algorithm, (ii) the separation matrix using iterative projection, and (iii) the dereverberation filters using multichannel linear prediction. Experimental results showed that the proposed method yielded higher separation performance and dereverberation performance than the baseline method under highly reverberant environments.

元の言語English
ホスト出版物のタイトル2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018 - Proceedings
出版者Institute of Electrical and Electronics Engineers Inc.
ページ31-35
ページ数5
2018-April
ISBN(印刷物)9781538646588
DOI
出版物ステータスPublished - 2018 9 10
イベント2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018 - Calgary, Canada
継続期間: 2018 4 152018 4 20

Other

Other2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018
Canada
Calgary
期間18/4/1518/4/20

Fingerprint

Factorization
Source separation
Reverberation
Microphones

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Electrical and Electronic Engineering

これを引用

Kagami, H., Kameoka, H., & Yukawa, M. (2018). Joint Separation and Dereverberation of Reverberant Mixtures with Determined Multichannel Non-Negative Matrix Factorization. : 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018 - Proceedings (巻 2018-April, pp. 31-35). [8462080] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/ICASSP.2018.8462080

Joint Separation and Dereverberation of Reverberant Mixtures with Determined Multichannel Non-Negative Matrix Factorization. / Kagami, Hideaki; Kameoka, Hirokazu; Yukawa, Masahiro.

2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018 - Proceedings. 巻 2018-April Institute of Electrical and Electronics Engineers Inc., 2018. p. 31-35 8462080.

研究成果: Conference contribution

Kagami, H, Kameoka, H & Yukawa, M 2018, Joint Separation and Dereverberation of Reverberant Mixtures with Determined Multichannel Non-Negative Matrix Factorization. : 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018 - Proceedings. 巻. 2018-April, 8462080, Institute of Electrical and Electronics Engineers Inc., pp. 31-35, 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018, Calgary, Canada, 18/4/15. https://doi.org/10.1109/ICASSP.2018.8462080
Kagami H, Kameoka H, Yukawa M. Joint Separation and Dereverberation of Reverberant Mixtures with Determined Multichannel Non-Negative Matrix Factorization. : 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018 - Proceedings. 巻 2018-April. Institute of Electrical and Electronics Engineers Inc. 2018. p. 31-35. 8462080 https://doi.org/10.1109/ICASSP.2018.8462080
Kagami, Hideaki ; Kameoka, Hirokazu ; Yukawa, Masahiro. / Joint Separation and Dereverberation of Reverberant Mixtures with Determined Multichannel Non-Negative Matrix Factorization. 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018 - Proceedings. 巻 2018-April Institute of Electrical and Electronics Engineers Inc., 2018. pp. 31-35
@inproceedings{3c490dba4ec742c89658e812af2c25d5,
title = "Joint Separation and Dereverberation of Reverberant Mixtures with Determined Multichannel Non-Negative Matrix Factorization",
abstract = "This paper proposes an extension of multichannel non-negative matrix factorization (MNMF) that simultaneously solves source separation and dereverberation. While MNMF was originally formulated under an underdetermined problem setting where sources can outnumber microphones, a determined counterpart of MNMF, which we call the determined MNMF (DMNMF), has recently been proposed with notable success. This approach is particularly notable in that the optimization process can be more than 30 times faster than the underdetermined version owing to the fact that it involves no matrix inversion computations. One drawback as regards all methods based on instantaneous mixture models, including MNMF, is that they are weak against long reverberation. To overcome this drawback, this paper proposes an extension of DMNMF using a frequency-domain convolutive mixture model. The optimization process of the proposed method consists of iteratively updating (i) the spectral parameters of each source using the majorization-minimization algorithm, (ii) the separation matrix using iterative projection, and (iii) the dereverberation filters using multichannel linear prediction. Experimental results showed that the proposed method yielded higher separation performance and dereverberation performance than the baseline method under highly reverberant environments.",
keywords = "Blind dereverberation, Blind source separation, Independent component analysis, Non-negative matrix factorization",
author = "Hideaki Kagami and Hirokazu Kameoka and Masahiro Yukawa",
year = "2018",
month = "9",
day = "10",
doi = "10.1109/ICASSP.2018.8462080",
language = "English",
isbn = "9781538646588",
volume = "2018-April",
pages = "31--35",
booktitle = "2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018 - Proceedings",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - GEN

T1 - Joint Separation and Dereverberation of Reverberant Mixtures with Determined Multichannel Non-Negative Matrix Factorization

AU - Kagami, Hideaki

AU - Kameoka, Hirokazu

AU - Yukawa, Masahiro

PY - 2018/9/10

Y1 - 2018/9/10

N2 - This paper proposes an extension of multichannel non-negative matrix factorization (MNMF) that simultaneously solves source separation and dereverberation. While MNMF was originally formulated under an underdetermined problem setting where sources can outnumber microphones, a determined counterpart of MNMF, which we call the determined MNMF (DMNMF), has recently been proposed with notable success. This approach is particularly notable in that the optimization process can be more than 30 times faster than the underdetermined version owing to the fact that it involves no matrix inversion computations. One drawback as regards all methods based on instantaneous mixture models, including MNMF, is that they are weak against long reverberation. To overcome this drawback, this paper proposes an extension of DMNMF using a frequency-domain convolutive mixture model. The optimization process of the proposed method consists of iteratively updating (i) the spectral parameters of each source using the majorization-minimization algorithm, (ii) the separation matrix using iterative projection, and (iii) the dereverberation filters using multichannel linear prediction. Experimental results showed that the proposed method yielded higher separation performance and dereverberation performance than the baseline method under highly reverberant environments.

AB - This paper proposes an extension of multichannel non-negative matrix factorization (MNMF) that simultaneously solves source separation and dereverberation. While MNMF was originally formulated under an underdetermined problem setting where sources can outnumber microphones, a determined counterpart of MNMF, which we call the determined MNMF (DMNMF), has recently been proposed with notable success. This approach is particularly notable in that the optimization process can be more than 30 times faster than the underdetermined version owing to the fact that it involves no matrix inversion computations. One drawback as regards all methods based on instantaneous mixture models, including MNMF, is that they are weak against long reverberation. To overcome this drawback, this paper proposes an extension of DMNMF using a frequency-domain convolutive mixture model. The optimization process of the proposed method consists of iteratively updating (i) the spectral parameters of each source using the majorization-minimization algorithm, (ii) the separation matrix using iterative projection, and (iii) the dereverberation filters using multichannel linear prediction. Experimental results showed that the proposed method yielded higher separation performance and dereverberation performance than the baseline method under highly reverberant environments.

KW - Blind dereverberation

KW - Blind source separation

KW - Independent component analysis

KW - Non-negative matrix factorization

UR - http://www.scopus.com/inward/record.url?scp=85054256334&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85054256334&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2018.8462080

DO - 10.1109/ICASSP.2018.8462080

M3 - Conference contribution

AN - SCOPUS:85054256334

SN - 9781538646588

VL - 2018-April

SP - 31

EP - 35

BT - 2018 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2018 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

ER -