AUDIO RECOGNITION DEVICE, AUDIO RECOGNITION METHOD, AND ELECTRONIC DEVICE

摘要

An audio characteristic amount calculation circuit (31) calculates an audio characteristic amount for each audio phrase inputted.  An estimation process likelihood calculation circuit (33) compares the calculated audio characteristic amount of each phrase to audio pattern sequence information on a plurality of phrases stored in a storage device (34) and calculates a plurality of candidates in the descending order of the likelihood value.  A recognition filtering device (4) judges whether to discard the extracted candidate in accordance with the likelihood difference ratio between the following two differences: a difference between the likelihood value of the first candidate and the likelihood value of the second candidate; and a difference between the likelihood value of the second candidate and the likelihood value of the third candidate.