摘要 |
<P>PROBLEM TO BE SOLVED: To provide a technology enabling high-accuracy performance of accent correction or prosody correction when synthesizing a speech. <P>SOLUTION: A speech synthesizer 1 comprises: a timing control part 11 for generating rhythm information corresponding to a mora quantity in text information 5; a rhythm information output part 12 for outputting the rhythm information to a speaker and a screen; a speech input part 13 for acquiring a first input speech synchronized with the rhythm information; a pitch extraction part 14 for extracting pitch frequency information of speech from the first input speech; a mora boundary correction part 15 for generating mora boundary information after correcting a mora boundary of the first input speech, from the rhythm information and the pitch frequency information; and an accent extraction part 16 for extracting accent information 6 from the text information 5, the mora boundary information of the first input speech and the pitch frequency information. <P>COPYRIGHT: (C)2012,JPO&INPIT |