Presentation 2012-07-20
Voice activity detection using density ratio estimation of speech and noise
Yuuki TACHIOKA, Toshiyuki HANAZAWA, Tomohiro NARITA, Jun ISHI,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In this paper, we propose a robust voice activity detection (VAD) method that uses a density ratio model. For VAD under highly noisy environments, the likelihood ratio test (LRT) is effective. Conventional LRT constructs speech and noise models, calculates the likelihood of each model, and takes the ratio of those likelihoods to detect speech. Here, there are two problems. First, in LRT, it is ignored that the likelihood ratio of speech and noise model is required, not the likelihood of each model. The proposed method directly estimates the likelihood ratio without calculating each likelihood using an obtained density ratio model. Second, there is the problem of determining thresholds, which are used for determining whether speech or not and significantly affect VAD performance. We propose a method that automatically determines thresholds using clustering analysis. The experiments show that the proposed method is more effective than conventional methods especially under non-stationary noisy environments, and that thresholds can be automatically determined according to noise features.
Keyword(in Japanese) (See Japanese page)
Keyword(in English) Voice activity detection / Likelihood ratio test / Density ratio estimation / Noise robustness / Automatic threshold determination
Paper # SP2012-54
Date of Issue

Conference Information
Committee SP
Conference Date 2012/7/12(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Speech (SP)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) Voice activity detection using density ratio estimation of speech and noise
Sub Title (in English)
Keyword(1) Voice activity detection
Keyword(2) Likelihood ratio test
Keyword(3) Density ratio estimation
Keyword(4) Noise robustness
Keyword(5) Automatic threshold determination
1st Author's Name Yuuki TACHIOKA
1st Author's Affiliation Information Technology R&D Center, Mitsubishi Electric Corporation()
2nd Author's Name Toshiyuki HANAZAWA
2nd Author's Affiliation Information Technology R&D Center, Mitsubishi Electric Corporation
3rd Author's Name Tomohiro NARITA
3rd Author's Affiliation Information Technology R&D Center, Mitsubishi Electric Corporation
4th Author's Name Jun ISHI
4th Author's Affiliation Information Technology R&D Center, Mitsubishi Electric Corporation
Date 2012-07-20
Paper # SP2012-54
Volume (vol) vol.112
Number (no) 141
Page pp.pp.-
#Pages 6
Date of Issue