| Presentation | 2009-06-25 A mean F_0 speaker adaptation method for regression model-based F_0 contour generation Hosana KAMIYAMA, Takahiro SHINOZAKI, Koji IWANO, Sadaoki FURUI, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | This paper proposes a new speaker adaptation method for the fundamental frequency (F_0) contour generation models based on the Quantification Theory (Type I). In this method, natural F_0 contour producing models for standard Japanese are trained using a large amount of speech data from many speakers, and natural as well as speaker-specific F_0 contours are generated by adapting mean F_0 values using a small amount of speech data from a specific speaker. Objective evaluation results using the models made by the proposed method confirm that around five sentences are enough for speaker adaptation. Subjective evaluation results confirm that naturalness of the synthesized speech using models adapted by 50 sentences is almost equivalent to that of the synthesized speech using models trained by 450 sentences for a specific speaker. These results indicate that the proposed adaptation method can produce highly natural synthesized speech. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | HMM-based Speech Synthesis / Quantification Theory (Type I) / F_0 Contour Generation / Prosody Control / Speaker Adaptation |
| Paper # | SP2009-38 |
| Date of Issue |
| Conference Information | |
| Committee | SP |
|---|---|
| Conference Date | 2009/6/17(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) | A mean F_0 speaker adaptation method for regression model-based F_0 contour generation |
| Sub Title (in English) | |
| Keyword(1) | HMM-based Speech Synthesis |
| Keyword(2) | Quantification Theory (Type I) |
| Keyword(3) | F_0 Contour Generation |
| Keyword(4) | Prosody Control |
| Keyword(5) | Speaker Adaptation |
| 1st Author's Name | Hosana KAMIYAMA |
| 1st Author's Affiliation | Department of Computer Science, Tokyo Institute of Technology() |
| 2nd Author's Name | Takahiro SHINOZAKI |
| 2nd Author's Affiliation | Department of Computer Science, Tokyo Institute of Technology |
| 3rd Author's Name | Koji IWANO |
| 3rd Author's Affiliation | Faculty of Environmental and Information Studies, Tokyo City University |
| 4th Author's Name | Sadaoki FURUI |
| 4th Author's Affiliation | Department of Computer Science, Tokyo Institute of Technology |
| Date | 2009-06-25 |
| Paper # | SP2009-38 |
| Volume (vol) | vol.109 |
| Number (no) | 99 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |