摘要 |
PROBLEM TO BE SOLVED: To provide a device and method for voice recognition which can reflect a recognition result characteristic of a user without new learning when the recognition result characteristic of the user is learnt at least once before use. SOLUTION: Text data in which pronunciation contents are described and voice data that the user speak corresponding to the text data are stored as a pair of pieces of data; and the text data and voice data are inputted and according to the text data and voice data as the pair of data, the recognition result characteristic of the user is learnt to generate a sound model or filter characteristic of the user. COPYRIGHT: (C)2003,JPO
|