摘要 |
PROBLEM TO BE SOLVED: To specify a recognition result, even when a speech recognition dictionary of large vocabulary is constituted of a plurality of recognition dictionaries. SOLUTION: A speech recognition device is constituted by dividing the speech recognition dictionary into a plurality of recognition dictionaries. Speech collected by a microphone 30 is stored in an internal memory (S100). By using a system which specifies a candidate of a high order of a recognition score by sequentially comparing magnitude relation of the recognition score, while a recognition score for expressing a recognition degree of each vocabulary of the recognition dictionary is calculated, speech is recognized for each recognition dictionary, and the candidate of the high order of the recognition score is specified (S106). The candidate of the high order of the recognition score is stored in a secondary determination dictionary (S108), and speech recognition of the speech stored in the internal memory is performed again by using the secondary determination dictionary (S114). COPYRIGHT: (C)2008,JPO&INPIT
|