Note: Your custom schedule will not be saved unless you create a new account or login to an existing account.
| Paper: | SP-P9.8 |
| Session: | Topics in Speech Synthesis |
| Time: | Wednesday, May 19, 15:30 - 17:30 |
| Presentation: |
Poster |
| Topic: |
Speech Processing: Speech Synthesis (including TTS) |
| Title: |
OPTIMIZING SUB-COST FUNCTIONS FOR SEGMENT SELECTION BASED ON PERCEPTUAL EVALUATIONS IN CONCATENATIVE SPEECH SYNTHESIS |
| Authors: |
Tomoki Toda; Nagoya Institute of Technology / ATR | | |
| | Hisashi Kawai; ATR, Spoken Language Translation Laboratories | | |
| | Minoru Tsuzaki; ATR, Spoken Language Translation Laboratories | | |
| Abstract: |
In concatenative speech synthesis, various factors affect thenaturalness of synthetic speech. A cost for segment selectionis calculated by integrating some sub-costs capturing thedegradation of naturalness caused by such factors. In thispaper, we optimize each sub-cost function for converting alinguistic feature or an acoustic parameter into a sub-costbased on perceptual evaluations. Two types of perceptualexperiments are performed with test sets constructed bycontrolling the variations of sub-costs to evaluate theindependent effect of each sub-cost and the interactionsbetween them. We clarify the effectiveness of perceptuallyoptimizing sub-cost functions from a result of a preferencetest comparing synthetic speech before and after the optimization. |
| |
| Back | |