MLSP 2007 data analysis competition: Frequency-domain blind source separation for convolutive mixtures of speech/audio signals

Hiroshi Sawada; Shoko Araki; Shoji Makino

doi:10.1109/MLSP.2007.4414280

MLSP 2007 data analysis competition: Frequency-domain blind source separation for convolutive mixtures of speech/audio signals

Hiroshi Sawada^*, Shoko Araki, Shoji Makino

^*Corresponding author for this work

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

15 Citations (Scopus)

Abstract

This paper describes the frequency-domain approach to the blind source separation of speech/audio signals that are convolutively mixed in a real room environment. With the application of shorttime Fourier transforms, convolutive mixtures in the time domain can be approximated as multiple instantaneous mixtures in the frequency domain. We employ complex-valued independent component analysis (ICA) to separate the mixtures in each frequency bin. Then, the permutation ambiguity of the ICA solutions should be aligned so that the separated signals are constructed properly in the time domain. We propose a permutation alignment method based on clustering the activity sequences of the frequency bin-wise separated signals. We achieved the overall winner status of MLSP 2007 Data Analysis Competition based on the presented method.

Original language	English
Title of host publication	Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP
Pages	45-50
Number of pages	6
DOIs	https://doi.org/10.1109/MLSP.2007.4414280
Publication status	Published - 2007
Externally published	Yes
Event	17th IEEE International Workshop on Machine Learning for Signal Processing, MLSP-2007 - Thessaloniki, Greece Duration: 2007 Aug 27 → 2007 Aug 29

Publication series

Name	Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP

Conference

Conference	17th IEEE International Workshop on Machine Learning for Signal Processing, MLSP-2007
Country/Territory	Greece
City	Thessaloniki
Period	07/8/27 → 07/8/29

ASJC Scopus subject areas

Computer Science(all)
Signal Processing

Access to Document

10.1109/MLSP.2007.4414280

Cite this

Sawada, H., Araki, S., & Makino, S. (2007). MLSP 2007 data analysis competition: Frequency-domain blind source separation for convolutive mixtures of speech/audio signals. In Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP (pp. 45-50). Article 4414280 (Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP). https://doi.org/10.1109/MLSP.2007.4414280

MLSP 2007 data analysis competition: Frequency-domain blind source separation for convolutive mixtures of speech/audio signals. / Sawada, Hiroshi; Araki, Shoko; Makino, Shoji.
Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP. 2007. p. 45-50 4414280 (Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP).

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

Sawada, H, Araki, S & Makino, S 2007, MLSP 2007 data analysis competition: Frequency-domain blind source separation for convolutive mixtures of speech/audio signals. in Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP., 4414280, Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP, pp. 45-50, 17th IEEE International Workshop on Machine Learning for Signal Processing, MLSP-2007, Thessaloniki, Greece, 07/8/27. https://doi.org/10.1109/MLSP.2007.4414280

Sawada H, Araki S, Makino S. MLSP 2007 data analysis competition: Frequency-domain blind source separation for convolutive mixtures of speech/audio signals. In Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP. 2007. p. 45-50. 4414280. (Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP). doi: 10.1109/MLSP.2007.4414280

Sawada, Hiroshi ; Araki, Shoko ; Makino, Shoji. / MLSP 2007 data analysis competition : Frequency-domain blind source separation for convolutive mixtures of speech/audio signals. Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP. 2007. pp. 45-50 (Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP).

@inproceedings{54c4868bd2cc41588e2d01fce319e507,

title = "MLSP 2007 data analysis competition: Frequency-domain blind source separation for convolutive mixtures of speech/audio signals",

abstract = "This paper describes the frequency-domain approach to the blind source separation of speech/audio signals that are convolutively mixed in a real room environment. With the application of shorttime Fourier transforms, convolutive mixtures in the time domain can be approximated as multiple instantaneous mixtures in the frequency domain. We employ complex-valued independent component analysis (ICA) to separate the mixtures in each frequency bin. Then, the permutation ambiguity of the ICA solutions should be aligned so that the separated signals are constructed properly in the time domain. We propose a permutation alignment method based on clustering the activity sequences of the frequency bin-wise separated signals. We achieved the overall winner status of MLSP 2007 Data Analysis Competition based on the presented method.",

author = "Hiroshi Sawada and Shoko Araki and Shoji Makino",

year = "2007",

doi = "10.1109/MLSP.2007.4414280",

language = "English",

isbn = "1424415667",

series = "Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP",

pages = "45--50",

booktitle = "Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP",

note = "17th IEEE International Workshop on Machine Learning for Signal Processing, MLSP-2007 ; Conference date: 27-08-2007 Through 29-08-2007",

}

TY - GEN

T1 - MLSP 2007 data analysis competition

T2 - 17th IEEE International Workshop on Machine Learning for Signal Processing, MLSP-2007

AU - Sawada, Hiroshi

AU - Araki, Shoko

AU - Makino, Shoji

PY - 2007

Y1 - 2007

N2 - This paper describes the frequency-domain approach to the blind source separation of speech/audio signals that are convolutively mixed in a real room environment. With the application of shorttime Fourier transforms, convolutive mixtures in the time domain can be approximated as multiple instantaneous mixtures in the frequency domain. We employ complex-valued independent component analysis (ICA) to separate the mixtures in each frequency bin. Then, the permutation ambiguity of the ICA solutions should be aligned so that the separated signals are constructed properly in the time domain. We propose a permutation alignment method based on clustering the activity sequences of the frequency bin-wise separated signals. We achieved the overall winner status of MLSP 2007 Data Analysis Competition based on the presented method.

AB - This paper describes the frequency-domain approach to the blind source separation of speech/audio signals that are convolutively mixed in a real room environment. With the application of shorttime Fourier transforms, convolutive mixtures in the time domain can be approximated as multiple instantaneous mixtures in the frequency domain. We employ complex-valued independent component analysis (ICA) to separate the mixtures in each frequency bin. Then, the permutation ambiguity of the ICA solutions should be aligned so that the separated signals are constructed properly in the time domain. We propose a permutation alignment method based on clustering the activity sequences of the frequency bin-wise separated signals. We achieved the overall winner status of MLSP 2007 Data Analysis Competition based on the presented method.

UR - http://www.scopus.com/inward/record.url?scp=48149093698&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=48149093698&partnerID=8YFLogxK

U2 - 10.1109/MLSP.2007.4414280

DO - 10.1109/MLSP.2007.4414280

M3 - Conference contribution

AN - SCOPUS:48149093698

SN - 1424415667

SN - 9781424415663

T3 - Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP

SP - 45

EP - 50

BT - Machine Learning for Signal Processing 17 - Proceedings of the 2007 IEEE Signal Processing Society Workshop, MLSP

Y2 - 27 August 2007 through 29 August 2007

ER -

MLSP 2007 data analysis competition: Frequency-domain blind source separation for convolutive mixtures of speech/audio signals

Abstract

Publication series

Conference

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this