Time domain blind source separation of non-stationary convolved signals by utilizing geometric beamforming

R. Aichner; S. Araki; S. Makino; T. Nishikawa; H. Saruwatari

doi:10.1109/NNSP.2002.1030056

Time domain blind source separation of non-stationary convolved signals by utilizing geometric beamforming

R. Aichner, S. Araki, S. Makino, T. Nishikawa, H. Saruwatari

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

29 Citations (Scopus)

Abstract

We propose a time-domain blind source separation (BSS) algorithm that utilizes geometric information such as sensor positions and assumed locations of sources. The algorithm tackles the problem of convolved mixtures by explicitly exploiting the non-stationarity of the acoustic sources. The learning rule is based on second-order statistics and is derived by natural gradient minimization. The proposed initialization of the algorithm is based on the null beamforming principle. This method leads to improved separation performance, and the algorithm is able to estimate long unmixing FIR filters in the time domain due to the geometric initialization. We also propose a post-filtering method for dewhitening which is based on the scaling technique in frequency-domain BSS. The validity of the proposed method is shown by computer simulations. Our experimental results confirm that the algorithm is capable of separating real-world speech mixtures and can be applied to short learning data sets down to a few seconds. Our results also confirm that the proposed dewhitening post-filtering method maintains the spectral content of the original speech in the separated output.

Original language	English
Title of host publication	Neural Networks for Signal Processing XII - Proceedings of the 2002 IEEE Signal Processing Society Workshop, NNSP 2002
Editors	Samy Bengio, Scott Douglas, Tulay Adali, Jan Larsen, Herve Bourlard
Publisher	Institute of Electrical and Electronics Engineers Inc.
Pages	445-454
Number of pages	10
ISBN (Electronic)	0780376161
DOIs	https://doi.org/10.1109/NNSP.2002.1030056
Publication status	Published - 2002
Externally published	Yes
Event	12th IEEE Workshop on Neural Networks for Signal Processing, NNSP 2002 - Martigny, Switzerland Duration: 2002 Sept 6 → …

Publication series

Name	Neural Networks for Signal Processing - Proceedings of the IEEE Workshop
Volume	2002-January

Other

Other	12th IEEE Workshop on Neural Networks for Signal Processing, NNSP 2002
Country/Territory	Switzerland
City	Martigny
Period	02/9/6 → …

ASJC Scopus subject areas

Electrical and Electronic Engineering
Artificial Intelligence
Software
Computer Networks and Communications
Signal Processing

Access to Document

10.1109/NNSP.2002.1030056

Cite this

Aichner, R., Araki, S., Makino, S., Nishikawa, T., & Saruwatari, H. (2002). Time domain blind source separation of non-stationary convolved signals by utilizing geometric beamforming. In S. Bengio, S. Douglas, T. Adali, J. Larsen, & H. Bourlard (Eds.), Neural Networks for Signal Processing XII - Proceedings of the 2002 IEEE Signal Processing Society Workshop, NNSP 2002 (pp. 445-454). Article 1030056 (Neural Networks for Signal Processing - Proceedings of the IEEE Workshop; Vol. 2002-January). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/NNSP.2002.1030056

Time domain blind source separation of non-stationary convolved signals by utilizing geometric beamforming. / Aichner, R.; Araki, S.; Makino, S. et al.
Neural Networks for Signal Processing XII - Proceedings of the 2002 IEEE Signal Processing Society Workshop, NNSP 2002. ed. / Samy Bengio; Scott Douglas; Tulay Adali; Jan Larsen; Herve Bourlard. Institute of Electrical and Electronics Engineers Inc., 2002. p. 445-454 1030056 (Neural Networks for Signal Processing - Proceedings of the IEEE Workshop; Vol. 2002-January).

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

Aichner, R, Araki, S, Makino, S, Nishikawa, T & Saruwatari, H 2002, Time domain blind source separation of non-stationary convolved signals by utilizing geometric beamforming. in S Bengio, S Douglas, T Adali, J Larsen & H Bourlard (eds), Neural Networks for Signal Processing XII - Proceedings of the 2002 IEEE Signal Processing Society Workshop, NNSP 2002., 1030056, Neural Networks for Signal Processing - Proceedings of the IEEE Workshop, vol. 2002-January, Institute of Electrical and Electronics Engineers Inc., pp. 445-454, 12th IEEE Workshop on Neural Networks for Signal Processing, NNSP 2002, Martigny, Switzerland, 02/9/6. https://doi.org/10.1109/NNSP.2002.1030056

Aichner R, Araki S, Makino S, Nishikawa T, Saruwatari H. Time domain blind source separation of non-stationary convolved signals by utilizing geometric beamforming. In Bengio S, Douglas S, Adali T, Larsen J, Bourlard H, editors, Neural Networks for Signal Processing XII - Proceedings of the 2002 IEEE Signal Processing Society Workshop, NNSP 2002. Institute of Electrical and Electronics Engineers Inc. 2002. p. 445-454. 1030056. (Neural Networks for Signal Processing - Proceedings of the IEEE Workshop). doi: 10.1109/NNSP.2002.1030056

Aichner, R. ; Araki, S. ; Makino, S. et al. / Time domain blind source separation of non-stationary convolved signals by utilizing geometric beamforming. Neural Networks for Signal Processing XII - Proceedings of the 2002 IEEE Signal Processing Society Workshop, NNSP 2002. editor / Samy Bengio ; Scott Douglas ; Tulay Adali ; Jan Larsen ; Herve Bourlard. Institute of Electrical and Electronics Engineers Inc., 2002. pp. 445-454 (Neural Networks for Signal Processing - Proceedings of the IEEE Workshop).

@inproceedings{f3cd75a1faff47f8909230cd3cb12744,

title = "Time domain blind source separation of non-stationary convolved signals by utilizing geometric beamforming",

abstract = "We propose a time-domain blind source separation (BSS) algorithm that utilizes geometric information such as sensor positions and assumed locations of sources. The algorithm tackles the problem of convolved mixtures by explicitly exploiting the non-stationarity of the acoustic sources. The learning rule is based on second-order statistics and is derived by natural gradient minimization. The proposed initialization of the algorithm is based on the null beamforming principle. This method leads to improved separation performance, and the algorithm is able to estimate long unmixing FIR filters in the time domain due to the geometric initialization. We also propose a post-filtering method for dewhitening which is based on the scaling technique in frequency-domain BSS. The validity of the proposed method is shown by computer simulations. Our experimental results confirm that the algorithm is capable of separating real-world speech mixtures and can be applied to short learning data sets down to a few seconds. Our results also confirm that the proposed dewhitening post-filtering method maintains the spectral content of the original speech in the separated output.",

author = "R. Aichner and S. Araki and S. Makino and T. Nishikawa and H. Saruwatari",

note = "Publisher Copyright: {\textcopyright} 2002 IEEE.; 12th IEEE Workshop on Neural Networks for Signal Processing, NNSP 2002 ; Conference date: 06-09-2002",

year = "2002",

doi = "10.1109/NNSP.2002.1030056",

language = "English",

series = "Neural Networks for Signal Processing - Proceedings of the IEEE Workshop",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

pages = "445--454",

editor = "Samy Bengio and Scott Douglas and Tulay Adali and Jan Larsen and Herve Bourlard",

booktitle = "Neural Networks for Signal Processing XII - Proceedings of the 2002 IEEE Signal Processing Society Workshop, NNSP 2002",

}

TY - GEN

T1 - Time domain blind source separation of non-stationary convolved signals by utilizing geometric beamforming

AU - Aichner, R.

AU - Araki, S.

AU - Makino, S.

AU - Nishikawa, T.

AU - Saruwatari, H.

PY - 2002

Y1 - 2002

N2 - We propose a time-domain blind source separation (BSS) algorithm that utilizes geometric information such as sensor positions and assumed locations of sources. The algorithm tackles the problem of convolved mixtures by explicitly exploiting the non-stationarity of the acoustic sources. The learning rule is based on second-order statistics and is derived by natural gradient minimization. The proposed initialization of the algorithm is based on the null beamforming principle. This method leads to improved separation performance, and the algorithm is able to estimate long unmixing FIR filters in the time domain due to the geometric initialization. We also propose a post-filtering method for dewhitening which is based on the scaling technique in frequency-domain BSS. The validity of the proposed method is shown by computer simulations. Our experimental results confirm that the algorithm is capable of separating real-world speech mixtures and can be applied to short learning data sets down to a few seconds. Our results also confirm that the proposed dewhitening post-filtering method maintains the spectral content of the original speech in the separated output.

AB - We propose a time-domain blind source separation (BSS) algorithm that utilizes geometric information such as sensor positions and assumed locations of sources. The algorithm tackles the problem of convolved mixtures by explicitly exploiting the non-stationarity of the acoustic sources. The learning rule is based on second-order statistics and is derived by natural gradient minimization. The proposed initialization of the algorithm is based on the null beamforming principle. This method leads to improved separation performance, and the algorithm is able to estimate long unmixing FIR filters in the time domain due to the geometric initialization. We also propose a post-filtering method for dewhitening which is based on the scaling technique in frequency-domain BSS. The validity of the proposed method is shown by computer simulations. Our experimental results confirm that the algorithm is capable of separating real-world speech mixtures and can be applied to short learning data sets down to a few seconds. Our results also confirm that the proposed dewhitening post-filtering method maintains the spectral content of the original speech in the separated output.

UR - http://www.scopus.com/inward/record.url?scp=84953720056&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84953720056&partnerID=8YFLogxK

U2 - 10.1109/NNSP.2002.1030056

DO - 10.1109/NNSP.2002.1030056

M3 - Conference contribution

AN - SCOPUS:84953720056

T3 - Neural Networks for Signal Processing - Proceedings of the IEEE Workshop

SP - 445

EP - 454

BT - Neural Networks for Signal Processing XII - Proceedings of the 2002 IEEE Signal Processing Society Workshop, NNSP 2002

A2 - Bengio, Samy

A2 - Douglas, Scott

A2 - Adali, Tulay

A2 - Larsen, Jan

A2 - Bourlard, Herve

PB - Institute of Electrical and Electronics Engineers Inc.

T2 - 12th IEEE Workshop on Neural Networks for Signal Processing, NNSP 2002

Y2 - 6 September 2002

ER -

Time domain blind source separation of non-stationary convolved signals by utilizing geometric beamforming

Abstract

Publication series

Other

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this