| 摘要 |
PROBLEM TO BE SOLVED: To improve the precision of voice recognition, specially, the recognition precision of characters, words, etc., given plural readings. SOLUTION: The same person tends to keep the same reading in the same conversation, namely, pronounce 'siti' as '7' consistently in the same conversation. Recognition probability corresponding to a reading that the person does not use in the case of answering in the conversation for the 1st time is decreased and recognition for 2nd and succeeding times is carried out. When the system repeats a recognition result through voice synthesis, the repetition is performed based upon the reading of the speaker having been recognized. For example, when the speaker pronounces 'siti' as '7', the system pronounces 'siti' repeatedly.
|
