Presentation | 2004/12/14 SNR and sub-band SNR estimation based on Gaussian mixture modeling in the log power domain with application for speech enhancements DAT Tran HUY, Hiroshi FUJIMURA, Kazuya TAKEDA, Fumitada ITAKURA, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | This work presents a flexible blind SNR estimation method based oh Gaussian mixture modeling (GMM) in the log-power domain. Considering the local noise and noisy speech powers as two log-normal distributed random variables, their distribution parameters are estimated via the EM algorithm and used to derive the segmental SNR, which is defined as the expectation distance between two subspace distributions. A compensation mode for the estimation under low SNR conditions ais also proposed. The experimental results, evaluated on the AURORA2 database, show the more consistency of proposed estimation method in both the noise conditions, compared to the conventional methods. The second application presented in this work is sub-band SNRs estimations for speech enhancement systems. Here the GMM is applied to each frequency bin, and then two methods of the sub-band SNRs estimation are proposed by using the. maximum a posterior (MAP) decomposition and cumulative distribution function (CDF) equalization. Furthermore, the sub-band SNR is used for the Wiener filtering systems. The evaluation experiments demonstrate the improvements of the proposed speech enhancement method in both segmental SNR and ASR performances. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | Gaussian mixture modeling / Signal to noise ratio / sub-band SNR / Wiener filtering |
Paper # | NLC2004-59,SP2004-99 |
Date of Issue |
Conference Information | |
Committee | NLC |
---|---|
Conference Date | 2004/12/14(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 | Natural Language Understanding and Models of Communication (NLC) |
---|---|
Language | ENG |
Title (in Japanese) | (See Japanese page) |
Sub Title (in Japanese) | (See Japanese page) |
Title (in English) | SNR and sub-band SNR estimation based on Gaussian mixture modeling in the log power domain with application for speech enhancements |
Sub Title (in English) | |
Keyword(1) | Gaussian mixture modeling |
Keyword(2) | Signal to noise ratio |
Keyword(3) | sub-band SNR |
Keyword(4) | Wiener filtering |
1st Author's Name | DAT Tran HUY |
1st Author's Affiliation | Graduate School of Information Science, Nagoya University() |
2nd Author's Name | Hiroshi FUJIMURA |
2nd Author's Affiliation | Graduate School of Information Science, Nagoya University |
3rd Author's Name | Kazuya TAKEDA |
3rd Author's Affiliation | Graduate School of Information Science, Nagoya University |
4th Author's Name | Fumitada ITAKURA |
4th Author's Affiliation | Graduate school of Information Engineering, Meijo University |
Date | 2004/12/14 |
Paper # | NLC2004-59,SP2004-99 |
Volume (vol) | vol.104 |
Number (no) | 539 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |