发明名称 |
SPEECH DATA CLASSIFICATION METHOD, SPEECH DATA CLASSIFICATION DEVICE, AND PROGRAM FOR THE SAME |
摘要 |
<P>PROBLEM TO BE SOLVED: To accurately classify speech data into speech phonemes automatically. <P>SOLUTION: Each classification time section classified by a phoneme boundary that is an estimation value of a boundary time between each of speech phonemes constituting a speech represented by speech data is classified by one or more state boundary so as to obtain plural state time sections. Each processing time section is formed by two adjacent classification time sections, and among a set of pairs of adjacent state time sections included in a processing time section of interest, a pair of adjacent state time sections having a maximum distance between representative values of speech data in both the adjacent state time sections is selected, and a state boundary or a phoneme boundary between the adjacent state time sections constituting the selected pair is set to be a phoneme boundary modified in the processing time section of interest. <P>COPYRIGHT: (C)2012,JPO&INPIT |
申请公布号 |
JP2011242497(A) |
申请公布日期 |
2011.12.01 |
申请号 |
JP20100112875 |
申请日期 |
2010.05.17 |
申请人 |
NIPPON TELEGR & TELEPH CORP <NTT> |
发明人 |
IJIMA YUSUKE;ISOGAI MITSUAKI;MIZUNO HIDEYUKI |
分类号 |
G10L15/04;G10L13/06 |
主分类号 |
G10L15/04 |
代理机构 |
|
代理人 |
|
主权项 |
|
地址 |
|