| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a speech synthesizer etc., for obtaining a natural synthesized speech fast or securely with simpler configuration as compared with a case wherein voice data whose reading matches the whole ideogram string are properly complemented. <P>SOLUTION: When data representing a fixed form message are supplied, a phrase editing section 5 divides a section representing wording meeting designated conditions among the data. Then a phrase database 7 is searched for phrase data of a phrase whose reading matches that of a phrase in the fixed form message after the division and a speaking speed is converted. The phrase editing section 5 also predicts the meter of the fixed form message meanwhile to select phrases matching respective phrases in the fixed form message most, one by one, out of searched phrase data according to the meter prediction result. For phrases which can not be selected, a sound processing section 41 is made to supply waveform data representing waveforms by unit speech. Then the selected phrase data and the waveform data that the sound processing section 41 is made to supply are combined with each other to generate data representing the synthesized speech. <P>COPYRIGHT: (C)2006,JPO&NCIPI |
