Font Size: a A A

The Design And Research Of Level-Testing System Of The Disyllable Of Mandarin Chinese

Posted on:2011-10-19Degree:MasterType:Thesis
Country:ChinaCandidate:H M LiFull Text:PDF
GTID:2178360305991887Subject:Communication and Information System
Abstract/Summary:PDF Full Text Request
With the development of the technology of computer and speech recognition, that makes the realiztion of Level-Testing System of Mandarin Chinese (LTSMC) become ture. LTSMC utilizes the computer to solve the limitation of manual. It is useful to unify the test standard, overcomes the shortcoming of subjective awareness, makes up the shortage of testers and popularizes the mandarin.This article mainly bases on the research of the speech recognizition especially the research of mandarin speech recogniziton to accomplish the design and realization of the Level-Testing System of the Disyllable of Mandarin Chinese (LTSDMC).The system involves in the collection of speech signal, pre-processing, feature extraction, the training and matching of the model as well as the scoring mode and so on.The test's main research works and achievements are as follows:Firstly, analyze the various kinds of theory and techonogy which are used and structure of the general speech recognition system, discuss the speech recognition system's basic principles and conceptions; including the teconology of the characteristic parameter extracting, the selection of speech recognition units, technology of training of model and template matching, etc.Secondly, systematically discuss the methods that extract the different main characteristic parameters of speech recognition, especially analyze and probe into the methods of extracting LPCC and MFCC parameters, and carry them through the emulate experiment.Thirdly, probe into the basic principle and the theoretic foundation and disposal method of recogniztion algorithm, and analyze the theory basement and the realizableness of the mathematic model according to the practical needs which the system adopts, and also introduce the specific realization steps.Finally, analyze scoring of the similar system and combine with the manual test to get the scoring mode of our system. On the platform of Microsoft Visual Studio 2008 and MATLAB 6.5, I utilize the language VC++ and Matlab to realize the LTSDM, and test it.The system mainly tests the phrases. The experimental test shows that this system can accomplish the expected features. By comparing the same person's twice test results and the test results of different people to determine the effectiveness of the system.
Keywords/Search Tags:speech recognition, feature extraction, template matching, Linear Prediction Cestrum Coefficient, Mel Frequency Cepstrum Coefficient
PDF Full Text Request
Related items