METHOD AND APPARATUS FOR SPEECH RECOGNITION USING RELIABILITY OF ARTICULATORY FEATURE

摘要

A method and an apparatus for speech recognition using reliability of an articulatory feature are provided to improve capability of the speech recognition. A method for speech recognition using reliability of an articulatory feature includes: receiving an audio signal and extracting an audio feature; extracting at least one of articulatory features corresponding to the audio feature which is extracted; calculating scores of the articulatory features per a phoneme; and calculating an observation score corresponding to the articulatory feature score per the phoneme and the observation probability value of the audio feature which is extracted. The articulatory feature per the phoneme searches at least one of the phoneme corresponding to the articulatory feature and calculates reliability of the articulatory features.