摘要 |
A speech acquisition unit (1) acquires an original speech uttered freely by a user. A speech data processing unit (7) processes an original speech signal to generate a processed speech signal. On the basis of a recognition score for each language calculated by a speech recognition unit (5) by recognizing and processing time-series data of acoustic features of the processed speech signal using the acoustic models (3-1 to 3-x) for each language, an acoustic model switching unit (4) determines a single acoustic model from among a plurality of acoustic models (3-1 to 3-x). |