| Presentation | 2005/8/19 A study on continuous-unit-first search algorithm for concatenative speech synthesis Nobuyuki NISHIZAWA, Hisashi KAWAI, |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In concatenative speech synthesis, discontinuities between speech segments degrade naturalness of synthetic sounds. Only when two segments are continuous in the original speech corpus, no degradation will be caused by concatenation. Therefore, search algorithm where continuous segment pairs in the original corpus are investigated by priority can be effective in unit selection. The algorithm is called "continuous-unit-first search" in this paper. We have already adopted the algorithm to our unit selection algorithm, which is based on breadth-first search. To investigate effects of continuous-unit-first search in unit selection, some experiments for comparison between search with and without the algorithm were conducted. In most cases, the results indicated effectiveness of the algorithm in cost value. But, in short-latency unit selection, when search depth is shallower than 300 milliseconds, the algorithm degrades results of unit selection. On the other hand, the algorithm may reduce computational effort for calculation of concatenation costs. However, result of experiment showed that reduction of computational effort was unimportant. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | concatenative speech synthesis / unit selection / dynamic programming / beam search / continuous-unit-first search |
| Paper # | SP2005-54 |
| Date of Issue |
| Conference Information | |
| Committee | SP |
|---|---|
| Conference Date | 2005/8/19(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 study on continuous-unit-first search algorithm for concatenative speech synthesis |
| Sub Title (in English) | |
| Keyword(1) | concatenative speech synthesis |
| Keyword(2) | unit selection |
| Keyword(3) | dynamic programming |
| Keyword(4) | beam search |
| Keyword(5) | continuous-unit-first search |
| 1st Author's Name | Nobuyuki NISHIZAWA |
| 1st Author's Affiliation | ATR Spoken Language Communication Research Laboratories() |
| 2nd Author's Name | Hisashi KAWAI |
| 2nd Author's Affiliation | ATR Spoken Language Communication Research Laboratories:KDDI R&D Laboratories Inc. |
| Date | 2005/8/19 |
| Paper # | SP2005-54 |
| Volume (vol) | vol.105 |
| Number (no) | 253 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |