A Codebook Adaptation Algorithm for SCHMM Using Formant Distribution
Download:
by Tae Young Yang
http://www.asel.udel.edu/icslp/cdrom/vol2/923/a923.pdf
Add To MetaCart
Abstract:
Won Ho Shin 1 This paper describes a codebook adaptation process improving the performance of speaker adaptation. The proposed method is performed prior to Bayesian speaker adaptation method using the formant distribution of adaptation data. The reference codebook is adapted to represent the formant distribution of a new speaker. The average recognition rate of Bayesian adaptation is improved from 91.4 � to 95.1 � using the proposed method. The proposed method is e�ective particularly when there exists a large mismatch between the reference codebook and a target speaker in feature space. In this cases the average recognition rate is 95.0 � while 89.9 � is obtained when only Bayesian adaptation is performed. 1.
Citations
| 26 | On the use of autocorrelation analysis for pitch detection – Rabiner - 1977 |
| 21 | Speaker adaptation based on MAP estimation of HMM parameters – Lee, Gauvain |
| 16 | A study on speaker adaptation of continuous density HMM parameters – Lee, Lin, et al. - 1990 |
| 7 | Unsupervised speaker adaptation method based on hierarchical spectral clustering – Furui - 1989 |
| 6 | A Comparative Study of Spectral Mapping for Speaker Adaptation – Nakamura, Shikano - 1990 |
| 3 | Fast Speaker Adaptation Combined with Soft Vector Quantization in an HMM Speech Recognition System – Class, Kaltenmeier, et al. - 1992 |
| 3 | A Piecewise Linear Spectral Mapping for Supervised Speaker Adaptation – Matsukoto, Inoue - 1992 |
| 1 | Speaker Adaptation Using Improved Speaker Markov Models – Rigoll |
| 1 | Frequency Domain Adaptive Post�ltering for Enhancement of Noisy Speech – Wang, Kabal, et al. - 1993 |
