| 摘要 |
PROBLEM TO BE SOLVED: To suppress the increase of data capacity of voice phonemic pieces and obtain synthesized voices having a plurality of voice qualities. SOLUTION: A voice phonemic piece database 37 is commonly provided between vector codebooks 40, 41 classified by phenome. A switch 39 switches to one of the vector codebooks 40, 41 classified by phenome in response to external designation for voice quality. A phenomic piece selector 34 refers to the phenomic piece database 37, controls a voice characteristic vector selector 38, sequentially selects codebook in the switched vector codebook classified by phenome based on a phenomic sequence of a voice phenomic piece VCV, reads a corresponding voice characteristic vector based on an index obtained from the phenomic piece database 37, and transmits it to a voice synthesizer 35. Since only the vector codebooks classified by phenome are plurally possessed, the increase of a data quantity of the voice phenomic pieces is suppressed when voice synthesis is extended to a plurality of the voice qualities. |
