发明名称 |
VOICE CONVERSION FUNCTION LEARNING DEVICE, VOICE CONVERTER, VOICE CONVERSION FUNCTION LEARNING METHOD, VOICE CONVERSION METHOD AND PROGRAM |
摘要 |
PROBLEM TO BE SOLVED: To convert voice of an utterer into voice easy for a listener to understand while maintaining voice quality of the utterer.SOLUTION: A first speaker model learning part 110 learns voice uttered by multiple first speakers and generates a first speakers' average voice model. A second speaker model learning part 115 learns voice uttered by multiple second speakers and generates a second speakers' average voice model. A first speakers' voice synthesizing part 120 uses the first speakers' average voice model to generate synthetic sound of the first speakers' average voice. A second speakers' voice synthesizing part 125 uses the second speakers' average voice model to generate synthetic sound of the second speakers' average voice. A conversion function learning part 130 learns a first voice conversion function for converting the voice uttered by the first speakers into voice similar to the voice uttered by the second speakers, and also learns a second voice conversion function for converting the voice uttered by the second speakers into voice similar to the voice uttered by the first speakers. |
申请公布号 |
JP2013238819(A) |
申请公布日期 |
2013.11.28 |
申请号 |
JP20120113439 |
申请日期 |
2012.05.17 |
申请人 |
NIPPON TELEGR & TELEPH CORP <NTT> |
发明人 |
MIZUNO HIDEYUKI;IJIMA YUSUKE |
分类号 |
G10L21/007;G10L21/057 |
主分类号 |
G10L21/007 |
代理机构 |
|
代理人 |
|
主权项 |
|
地址 |
|