Process that is implemented by a computer and computer systems for automatic speech recognition based on the phonology

摘要

(57) [Abstract] The present invention linguistic knowledge, based on the use of phonological knowledge, particularly to guide the speech recognition process. Audio signal containing speech are received, linguistic cues in the speech signal is detected. From linguistic cues are detected, the display of the sound representation of the content of the audio signal is generated. The display of this sound representation, and a word segmentation part of at least one, set on each word split part - and a pair of rhyme, consisting of phonological elements associated. Phonological These elements are monovalent, and can be interpreted directly in the audio signal appears in all languages, and can be distinguished from each other. Dictionary of display of sound representation of predetermined is given to the word in a particular language. The recognized words best match to the display representation of the sound that is generated is found in the dictionary, uttered by this.