摘要 |
PROBLEM TO BE SOLVED: To provide a voice recognition device and voice recognition method with which parameter training of a recognition system can be carried out at high speed even when the operation environment changes and highly accurate voice recognition can be carried out, and to provide a recording medium with the voice recognition program recorded thereon. SOLUTION: In the voice recognition device, the pattern obtained by analyzing voice data is converted using an adjusting parameter, and the converted pattern is recognized as a feature pattern. In adjusting the adjusting parameter, the feature pattern using the voice data in a voice database is compared by a recognition model, and the adjustment parameter is so adjusted as to minimize the loss of the comparison result. COPYRIGHT: (C)2003,JPO
|