Presentation | 2000/9/22 Speaker Recognition Using Multi-SNR Subband GMM under Noisy Environments Kenichi YOSHIDA, Kazuyuki TAKAGI, Kazuhiko OZEKI, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | When a speech signal is divided into multiple subbands in the spectral domain, the contribution of each subband to speaker recognition is different, depending on the amount of speaker information and phonetic information contained in it. To exploit such a fact, a subband model is effective, which is a set of GMMs, each trained on a subband separately. In real environments, however, a speech signal is usually corrupted with noise. Therefore, in making such a model, it is necessary to take the effect of noise into consideration. In this paper, we propose a Multi-SNR Subband Model, which has enhanced rebustness against noise by using multiple GMMs trained with speech signals of various SNRs for each subband. In text-independent speaker indentification experiments under various noisy environments, it was confirmed that this model achieves a higher recognition performance than conventional models. It was also made clear that optimization of the recombination weights is effective to improve the performance. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | multi-SNR / subband GMM / speaker indentification / text-independence / likelihood recombination weght |
Paper # | DSP2000-97,SP2000-63 |
Date of Issue |
Conference Information | |
Committee | SP |
---|---|
Conference Date | 2000/9/22(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) | Speaker Recognition Using Multi-SNR Subband GMM under Noisy Environments |
Sub Title (in English) | |
Keyword(1) | multi-SNR |
Keyword(2) | subband GMM |
Keyword(3) | speaker indentification |
Keyword(4) | text-independence |
Keyword(5) | likelihood recombination weght |
1st Author's Name | Kenichi YOSHIDA |
1st Author's Affiliation | The University of Electro-Communications() |
2nd Author's Name | Kazuyuki TAKAGI |
2nd Author's Affiliation | The University of Electro-Communications |
3rd Author's Name | Kazuhiko OZEKI |
3rd Author's Affiliation | The University of Electro-Communications |
Date | 2000/9/22 |
Paper # | DSP2000-97,SP2000-63 |
Volume (vol) | vol.100 |
Number (no) | 328 |
Page | pp.pp.- |
#Pages | 7 |
Date of Issue |