| A Combined Speaker Adaptation Method for Mandarin Speech Recognition |
| |
| 作者姓名： | 徐向华 朱杰 |
| |
| 作者单位： | Dept.ofElectronicEng.,ShanghaiJiaotongUniv.,Shanghai200030,China |
| |
| 基金项目： | ScienceandTechnologyCommitteeofShanghai(01JC14033) |
| |
| 摘 要： | A speaker adaptation method that combines transformation matrix linear interpolation with maximum a posteriori (MAP) was proposed. Firstly this method can keep the asymptotical characteristic of MAP. Secondly,as the method uses linear interpolation with several speaker-dependent (SD) transformation matrixes, it can fully use the prior knowledge and keep fast adaptation. The experimental results show that the combined method achieves an 8.24% word error rate reduction with only one adaptation utterance, and keeps asymptotic to the performance of SD model for large amounts of adaotation data.
|
| 关 键 词： | 语音识别 MAP MLMI 讲话者适应 |
A Combined Speaker Adaptation Method for Mandarin Speech Recognition |
| |
| XU Xiang-hua,ZHU Jie.A Combined Speaker Adaptation Method for Mandarin Speech Recognition[J].Journal of Shanghai Jiaotong university,2004,9(4):21-24. |
| |
| Authors: | XU Xiang-hua ZHU Jie |
| |
| Institution: | Dept. of Electronic Eng., Shanghai Jiaotong Univ., Shanghai 200030, China |
| |
| Abstract: | A speaker adaptation method that combines transformation matrix linear interpolation with maximum a posteriori (MAP) was proposed. Firstly this method can keep the asymptotical characteristic of MAP. Secondly, as the method uses linear interpolation with several speaker-dependent (SD) transformation matrixes, it can fully use the prior knowledge and keep fast adaptation. The experimental results show that the combined method achieves an 8.24% word error rate reduction with only one adaptation utterance, and keeps asymptotic to the performance of SD model for large amounts of adaptation data. |
| |
| Keywords: | speech recognition speaker adaptation maximum a posteriori (MAP) maximum likelihood model interpolation (MLMI) |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
