Presentation | 2012-07-21 Acoustic model adaptation choosing static and dynamic streams in noisy environments Satoshi TAMURA, Satoru HAYAMIZU, |
---|---|
PDF Download Page | PDF download Page Link |
Abstract(in Japanese) | (See Japanese page) |
Abstract(in English) | In this paper, an acoustic model adaptation method based on multi streams is proposed for speech recognition in noisy or real environments. At first, an acoustic feature vector is divided into several vectors (e.g. static, first-order and second-order dynamic vectors), namely streams. Second, the order of streams is determined according to accuracies of pre-recognition results. While adaptation, a stream performing the highest recognition performance is updated for the stream only. Alternatively, a stream is adapted using all the streams that are superior to the stream. In order to evaluate the proposed technique, recognition and adaptation experiments were conducted using a corpus CENSREC-1. Pre-recognition results show dynamic features are more robust to noise than static parameters. And through adaptation experiments it is found that our proposed method achieved the best performance compared to conventional acoustic feature and its streams. These results show the effectiveness of our proposed adaptation scheme. |
Keyword(in Japanese) | (See Japanese page) |
Keyword(in English) | speech recognition in noisy condition / multi stream / model adaptation / MAP / MLLR / CENSREC |
Paper # | SP2012-56 |
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) | Acoustic model adaptation choosing static and dynamic streams in noisy environments |
Sub Title (in English) | |
Keyword(1) | speech recognition in noisy condition |
Keyword(2) | multi stream |
Keyword(3) | model adaptation |
Keyword(4) | MAP |
Keyword(5) | MLLR |
Keyword(6) | CENSREC |
1st Author's Name | Satoshi TAMURA |
1st Author's Affiliation | Department of Information Science, Faculty of Engineering, Gifu University() |
2nd Author's Name | Satoru HAYAMIZU |
2nd Author's Affiliation | Department of Information Science, Faculty of Engineering, Gifu University |
Date | 2012-07-21 |
Paper # | SP2012-56 |
Volume (vol) | vol.112 |
Number (no) | 141 |
Page | pp.pp.- |
#Pages | 6 |
Date of Issue |