摘要 |
PROBLEM TO BE SOLVED: To enable a voice recognition device to perform a voice recognizing processing having high accuracy by obtaining a proper rejection threshold from the first voice input of the first time and also to recognize an uttered word correctly regardless of the fluctuation of uttering methods correctly and to reject an erroneously recognized word properly. SOLUTION: This device is provided with a collation part 5 which calculates a collation distance indicating the degree of similarity between a voice pattern and a standard pattern for every registered standard pattern and decides a standard pattern whose degree of similarity with a voice pattern is the highest as a first recognition result and a standard pattern whose degree of similarity with the voice pattern is secondly high as a second recognition result and a rejection deciding part 7 performing a rejection decision by comparing the difference between collation distances of the first recognition result and the second recognition result which are decided by the collation part 5 with a rejection threshold which is to be updated for every voice input in dependence on an SN ratio.
|