| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a voice processor which appropriately processes paralanguage information. <P>SOLUTION: A determination tree learning section 36 which is the voice processor, includes a learning voice corpus 30, an acoustic analysis section 74 which extracts acoustic featured values for every uttering unit included in the learning voice corpus 30, a statistical processing section 78 which collects statistical information related to the paralanguage information being perceived by a listener during play back for every prescribed uttering unit and a learning processing section 76 which receives the acoustic featured values extracted by the acoustic analysis section 74, makes the statistical information collected by the statistical processing section 78 as correct data and conducts the learning, that ouputs the statistical information being optimized for the acoustic featured values by mechanical learning, for a determination tree group 38. <P>COPYRIGHT: (C)2006,JPO&NCIPI |
