Effective Frame Selection for Blind Source Separation Based on Frequency Domain Independent Component Analysis

Yusuke MIZUNO; Kazunobu KONDO; Takanori NISHINO; Norihide KITAOKA; Kazuya TAKEDA

doi:10.1587/transfun.E97.A.784

Effective Frame Selection for Blind Source Separation Based on Frequency Domain Independent Component Analysis

Yusuke MIZUNO, Kazunobu KONDO, Takanori NISHINO, Norihide KITAOKA, Kazuya TAKEDA

Full Text Views

0

Share
Cite this

Summary :

Blind source separation is a technique that can separate sound sources without such information as source location, the number of sources, and the utterance content. Multi-channel source separation using many microphones separates signals with high accuracy, even if there are many sources. However, these methods have extremely high computational complexity, which must be reduced. In this paper, we propose a computational complexity reduction method for blind source separation based on frequency domain independent component analysis (FDICA) and examine temporal data that are effective for source separation. A frame with many sound sources is effective for FDICA source separation. We assume that a frame with a low kurtosis has many sound sources and preferentially select such frames. In our proposed method, we used the log power spectrum and the kurtosis of the magnitude distribution of the observed data as selection criteria and conducted source separation experiments using speech signals from twelve speakers. We evaluated the separation performances by the signal-to-interference ratio (SIR) improvement score. From our results, the SIR improvement score was 24.3dB when all the frames were used, and 23.3dB when the 300 frames selected by our criteria were used. These results clarified that our proposed selection criteria based on kurtosis and magnitude is effective. Furthermore, we significantly reduced the computational complexity because it is proportional to the number of selected frames.

Publication: IEICE TRANSACTIONS on Fundamentals Vol.E97-A No.3 pp.784-791

Publication Date: 2014/03/01

Publicized

Online ISSN: 1745-1337

DOI: 10.1587/transfun.E97.A.784

Type of Manuscript: PAPER

Category: Engineering Acoustics

Authors

Yusuke MIZUNO
  Nagoya University
Kazunobu KONDO
  Yamaha Corporation
Takanori NISHINO
  Mie University
Norihide KITAOKA
  Nagoya University
Kazuya TAKEDA
  Nagoya University

Keyword

dodecahedral microphone array, frequency domain independent component analysis, computational complexity reduction, signal to interference ratio improvement

Cite this

Copy

Yusuke MIZUNO, Kazunobu KONDO, Takanori NISHINO, Norihide KITAOKA, Kazuya TAKEDA, "Effective Frame Selection for Blind Source Separation Based on Frequency Domain Independent Component Analysis" in IEICE TRANSACTIONS on Fundamentals, vol. E97-A, no. 3, pp. 784-791, March 2014, doi: 10.1587/transfun.E97.A.784.
Abstract: Blind source separation is a technique that can separate sound sources without such information as source location, the number of sources, and the utterance content. Multi-channel source separation using many microphones separates signals with high accuracy, even if there are many sources. However, these methods have extremely high computational complexity, which must be reduced. In this paper, we propose a computational complexity reduction method for blind source separation based on frequency domain independent component analysis (FDICA) and examine temporal data that are effective for source separation. A frame with many sound sources is effective for FDICA source separation. We assume that a frame with a low kurtosis has many sound sources and preferentially select such frames. In our proposed method, we used the log power spectrum and the kurtosis of the magnitude distribution of the observed data as selection criteria and conducted source separation experiments using speech signals from twelve speakers. We evaluated the separation performances by the signal-to-interference ratio (SIR) improvement score. From our results, the SIR improvement score was 24.3dB when all the frames were used, and 23.3dB when the 300 frames selected by our criteria were used. These results clarified that our proposed selection criteria based on kurtosis and magnitude is effective. Furthermore, we significantly reduced the computational complexity because it is proportional to the number of selected frames.
URL: https://globals.ieice.org/en_transactions/fundamentals/10.1587/transfun.E97.A.784/_p

Copy

@ARTICLE{e97-a_3_784,
author={Yusuke MIZUNO, Kazunobu KONDO, Takanori NISHINO, Norihide KITAOKA, Kazuya TAKEDA, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Effective Frame Selection for Blind Source Separation Based on Frequency Domain Independent Component Analysis},
year={2014},
volume={E97-A},
number={3},
pages={784-791},
abstract={Blind source separation is a technique that can separate sound sources without such information as source location, the number of sources, and the utterance content. Multi-channel source separation using many microphones separates signals with high accuracy, even if there are many sources. However, these methods have extremely high computational complexity, which must be reduced. In this paper, we propose a computational complexity reduction method for blind source separation based on frequency domain independent component analysis (FDICA) and examine temporal data that are effective for source separation. A frame with many sound sources is effective for FDICA source separation. We assume that a frame with a low kurtosis has many sound sources and preferentially select such frames. In our proposed method, we used the log power spectrum and the kurtosis of the magnitude distribution of the observed data as selection criteria and conducted source separation experiments using speech signals from twelve speakers. We evaluated the separation performances by the signal-to-interference ratio (SIR) improvement score. From our results, the SIR improvement score was 24.3dB when all the frames were used, and 23.3dB when the 300 frames selected by our criteria were used. These results clarified that our proposed selection criteria based on kurtosis and magnitude is effective. Furthermore, we significantly reduced the computational complexity because it is proportional to the number of selected frames.},
keywords={},
doi={10.1587/transfun.E97.A.784},
ISSN={1745-1337},
month={March},}

Copy

TY - JOUR
TI - Effective Frame Selection for Blind Source Separation Based on Frequency Domain Independent Component Analysis
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 784
EP - 791
AU - Yusuke MIZUNO
AU - Kazunobu KONDO
AU - Takanori NISHINO
AU - Norihide KITAOKA
AU - Kazuya TAKEDA
PY - 2014
DO - 10.1587/transfun.E97.A.784
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E97-A
IS - 3
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - March 2014
AB - Blind source separation is a technique that can separate sound sources without such information as source location, the number of sources, and the utterance content. Multi-channel source separation using many microphones separates signals with high accuracy, even if there are many sources. However, these methods have extremely high computational complexity, which must be reduced. In this paper, we propose a computational complexity reduction method for blind source separation based on frequency domain independent component analysis (FDICA) and examine temporal data that are effective for source separation. A frame with many sound sources is effective for FDICA source separation. We assume that a frame with a low kurtosis has many sound sources and preferentially select such frames. In our proposed method, we used the log power spectrum and the kurtosis of the magnitude distribution of the observed data as selection criteria and conducted source separation experiments using speech signals from twelve speakers. We evaluated the separation performances by the signal-to-interference ratio (SIR) improvement score. From our results, the SIR improvement score was 24.3dB when all the frames were used, and 23.3dB when the 300 frames selected by our criteria were used. These results clarified that our proposed selection criteria based on kurtosis and magnitude is effective. Furthermore, we significantly reduced the computational complexity because it is proportional to the number of selected frames.
ER -