摘要 |
PROBLEM TO BE SOLVED: To provide a method and a device for element piece set generation that can decrease the number of elementary pieces used for speech synthesis, while suppressing deterioration in voice quality. SOLUTION: Element piece sets (506) before being updated are read in (S501) and clustered (S502), while a phonemic environment is taken into consideration. Then a representative element piece of an element piece set, belonging to each of clusters obtained by the clustering, is generated for each cluster (S503), and an element piece belonging to each cluster is replaced with a generated elementary piece for each cluster to update the elementary piece sets (506). COPYRIGHT: (C)2006,JPO&NCIPI
|