We first introduce the phylogeny and the actuality of speech recognition and the difficulties of speech recognition in this paper, and present a novel algorithm of speaker-independent continuous speech based on High-dimension Space Theory, which don't need syncopate and recognize directly. False recognition rate is vary low because of syncopate. In studying speech recognition, we first construct the speech model, the high-dimension space covering theory (HDSCT) is applied to the study process of the speaker-independent continuous greeting speech recognition, and the dynamic searching theory of high dimensional space vertex covering is done to dynamic search the rolling time axis of the continuous greeting speech. This algorithm has such advantages as simple, practicality, high speed, small EMS memory needed, and the online real time processes is applicable. This paper also describes the basic principle of HMM, the structure of HMM and the application in greeting speech recognition. At the same time it discusses relate HDSCT-based algorithms...
|