摘要 |
PROBLEM TO BE SOLVED: To provide a speech recognition apparatus which suppresses the number of recognition vocabularies and has efficient operability since a practical speech recognition apparatus needs to secure recognition precision, secure sufficient recognizable vocabularies, and facilitate input, although recognition precision becomes lower and the efficiency of input operation decreases as vocabularies to be recognized in speech recognition increase. SOLUTION: A storage device for speech recognition is constituted in a hierarchical structure and the inclusion range is made gradually narrower; and recognition processing is performed while next spoken contents are predicted. At the same time, upper layers are so corrected at a proper time that a long phrase can be inputted as the skill of a user is improved, and a vocabulary of low use frequency is replaced with a new vocabulary. COPYRIGHT: (C)2005,JPO&NCIPI
|