Font Size: a A A

Research And Implementation Of Speech Recognition System Based On Model-Matching

Posted on:2006-03-13Degree:MasterType:Thesis
Country:ChinaCandidate:Y F LvFull Text:PDF
GTID:2168360152991088Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
The speech recognition technology is the important developing direction of the computer technology, it has already become the key technology that the computer has popularized among hundreds of millions of common people, and will become significant technology of the information industry and important characteristic of the computer in the future. At present, the speech recognition technology has been already used widely, hundreds of speech recognition goods are produced every year. But speech recognition products such as voice-controlling car, voice-controlling telephone and voice-controlling electric home appliances etc. are few among people's daily life, this field has wide market prospects. Under this background, on the basis of studying the speech recognition technology overally and deeply, the studying of the word-isolated, speaker-dependent and small-scale vocabulary speech recognition system based on model-matching is put on more emphasis, improving and upgrading the existing research results, applying it to people's daily life, facilitating people's daily life and improving the quality of people's daily life.This paper firstly introduces the research and development of speech recognition in detail, has done a very good summary in such aspects as developing history, existing problems and developing direction of speech recognition, laiding a good foundation for further study of speech recognition. Then it discusses the basic implementation method of the speech recognition system in detail, especially laiding particular emphasis on the speech recognition system based on model-matching. Furthurly proposing one realization scheme that is suited to the word-isolated, speaker-dependent and small-scale vocabulary speech recognition system based on model-matching. Putting special attention to yawp-removing and the extreme-point-measuring etc., and improving DTW at four points that is important in the whole algorithm of the system. On this basis, preliminary realization to the system is carried on.
Keywords/Search Tags:model-matching, preprocessing, feature parameter extraction, dynamic time warping(DTW), measure-estimating, model-training
PDF Full Text Request
Related items