Presentation | 2014-12-16 Noise robust speech recognition by non-negative matrix factorization using GMM clustering in MFCC domain Kentaro FUJIGAKI, Yosuke KASHIWAGI, Daisuke SAITO, Nobuaki MINEMATSU, Keikichi HIROSE, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | Exemplar-based feature enhancement by non-negative matrix factorization (NMF) was proposed for noise-robust speech recognition. When we consider only additive noises, we can decompose a noisy speech spectrum into a linear but sparse combination of speech and noise bases. In the conventional NMF, decomposition is unsupervised. If we can give the phoneme sequence of an input utterance to the NMF processing, it is surely possible to realize much more precise decomposition. However, in the task of speech recognition, the phoneme sequence is unknown and unavailable. In this paper, therefore, we introduce unsupervised GMM clustering and classify each input frame by using GMM indexes. For NMF, speech bases are built separately for each GMM index. Experiments show that our proposed method of combining NMF with GMM clustering gives higher robustness of recognizing noisy speech than the original NMF. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | robust speech recognition / noise surpression / feature enhancement / NMF / GMM clustering |
Paper # | SP2014-113 |
Date of Issue |
Conference Information | |
Committee | SP |
---|---|
Conference Date | 2014/12/8(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) | Noise robust speech recognition by non-negative matrix factorization using GMM clustering in MFCC domain |
Sub Title (in English) | |
Keyword(1) | robust speech recognition |
Keyword(2) | noise surpression |
Keyword(3) | feature enhancement |
Keyword(4) | NMF |
Keyword(5) | GMM clustering |
1st Author's Name | Kentaro FUJIGAKI |
1st Author's Affiliation | The University of Tokyo() |
2nd Author's Name | Yosuke KASHIWAGI |
2nd Author's Affiliation | The University of Tokyo |
3rd Author's Name | Daisuke SAITO |
3rd Author's Affiliation | The University of Tokyo |
4th Author's Name | Nobuaki MINEMATSU |
4th Author's Affiliation | The University of Tokyo |
5th Author's Name | Keikichi HIROSE |
5th Author's Affiliation | The University of Tokyo |
Date | 2014-12-16 |
Paper # | SP2014-113 |
Volume (vol) | vol.114 |
Number (no) | 365 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |