Speech Evaluation System Based on MFCC

Abstract

In order to evaluate the speaker's pronunciation quality, a speech evaluation system based on MFCC is designed by using some techniques including pre-processing, feature extraction and pattern matching and so on. The MFCC parameters of the standard speech and those of the original speech are extracted. The result is given by the mechanism. Experimental results show that the scores assessed by the system is relating to people’s subjective feeling closely. The software is developed by VC++ and runs on windows XP.