How does speech recognition work?
Posted by Peter Maddern on 19th April 2009
Speech recognition as exemplified by Dragon NaturallySpeaking is a complex process but can be broken down to contributions from 3 models as follows:-
The “Acoustic model” represents the connection between language sounds (phonemes) and the voice signal.
The “Lexicon” represents the connection between phonemes and words.
The “Language model” represents the connection between words and phrases/sentences.
Peter Maddern
Posted in Speech recognition | No Comments »