| 摘要 |
PURPOSE:To accurately recognize a phoneme string by recognizing the string in consideration of an index which indicates whether there is probability that a phoneme is erroneously extracted or not with respect to the vowel or the consonant of each recognized phoneme or each phoneme constituting each recognized phoneme string. CONSTITUTION:A voice to be recognized is recognized by a first recognizing means, and the index indicating whether there is probability that the phoneme is erroneous recognized or not is added to at least the vowel or the consonant of each of phonemes of recognition results or each phoneme constituting the phoneme string of recognition results. A second recognizing means refers to this index to recognize the voice as a string of phonemes or the like. Thus, the voice is accurately recognized with the simple constitution, and a complicated trouble of re-input or the like for erroneous recognition is omitted. |
