| 摘要 |
PROBLEM TO BE SOLVED: To reduce the number of recognition failures by giving flexibility to read information for generating a syllable and a phoneme model string and their numbers, which are used for speech recognition processing, according to the number of syllables and the number of phonemes of a word in a speech recognition device. SOLUTION: The speech recognition device includes: a dictionary for speech recognition, which stores correspondence of the word and a variety of read information, and a fluctuation level for indicating its difference degree; a speech signal input section for receiving input of a speech signal; a read information selection section for selecting the read information which satisfies a predetermined condition regarding the fluctuation level; and a speech recognition section which performs speech recognition by using the syllable and the phoneme model string, and which determines whether, the speech signal corresponding to the word stored in the dictionary for speech recognition is included, to output the corresponding word as the speech recognition results, if included. COPYRIGHT: (C)2010,JPO&INPIT
|
