摘要 |
PROBLEM TO BE SOLVED: To shorten a speech recognition time. SOLUTION: When a speech is input, a syllable section of three syllables from the head of the input speech is cut out by using a syllable recognizing method etc., and the cut syllable section is speech-recognized (102); and dictionary identification information, corresponding to the speech-recognized syllable string of the three syllables, is acquired by referring to indexes for reference, and a small-vocabulary word dictionary database that the dictionary identification information indicates is selected as a small-vocabulary word dictionary database to be used for speech recognition processing (104). The selected small-vocabulary word dictionary database is read in and stored in a work memory (106), and all the vocabulary words stored in the small-vocabulary word dictionary database stored in the work memory are searched sequentially for to speech-recognize words that the input speech represents (108). COPYRIGHT: (C)2007,JPO&INPIT
|