发明名称 |
VOICE RECOGNITION METHOD AND APPARATUS FOR TWO-CHANNEL VOICE, AND PROGRAM |
摘要 |
<P>PROBLEM TO BE SOLVED: To improve a precision of voice recognition reliability at a voice transmitting side in two channels of transmitting and receiving sides. <P>SOLUTION: A voice recognition method of two-channel voices includes a voice recognition step and a recognition reliability calculation step. The voice recognition step includes inputting a transmitting-side voice and a receiving-side voice, and outputting a transmitting-side voice recognition result and a receiving-side voice recognition result in which word recognition reliability is imparted for each word obtained by performing voice recognition processing on the voices. The recognition reliability calculation step includes inputting the transmitting-side voice recognition result and the receiving-side voice recognition result, referring to a word relation degree table indicating a relation degree between two words obtained by combining all the words of the voice recognition results, determining an intra-channel context consistency scale of each voice recognition result and an inter-channel context consistency scale between the transmitting-side voice recognition result and the receiving-side voice recognition result, and calculating out a weighted sum of the transmitting-side intra-channel context consistency scale and inter-channel context consistency scale as transmitting-side recognition reliability. <P>COPYRIGHT: (C)2012,JPO&INPIT |
申请公布号 |
JP2012027065(A) |
申请公布日期 |
2012.02.09 |
申请号 |
JP20100162629 |
申请日期 |
2010.07.20 |
申请人 |
NIPPON TELEGR & TELEPH CORP <NTT> |
发明人 |
ASAMI TAICHI;NOMOTO NARIHISA;KOBASHIGAWA SATORU;MASATAKI HIROKAZU |
分类号 |
G10L15/10;G10L15/00;G10L15/18 |
主分类号 |
G10L15/10 |
代理机构 |
|
代理人 |
|
主权项 |
|
地址 |
|