| 摘要 |
<P>PROBLEM TO BE SOLVED: To reduce the number of phoneme data necessary for voice synthesis. <P>SOLUTION: A storage device 14 stores phoneme data DS indicating a voice sound of a speaker US for each voice phoneme. A voice quality conversion part 24 sequentially generates phoneme data DST corresponding to a voice of a speaker UT by applying to the phoneme data DS corresponding to a sound production character to be synthesized a conversion function F (X) that is generated from a mixed distribution model λ (Z) indicating a probability distribution between feature quantity information X of a voice of the speaker US and feature quantity information Y of the voice of the speaker UT to convert the voice of the speaker US into the voice of the speaker UT. A voice synthesis part 26 generates a voice signal VSYN from each phoneme data DST generated from the voice quality conversion part 24. <P>COPYRIGHT: (C)2012,JPO&INPIT |
