A study on the automatic pronunciation correction in english learning

Abstract

In this thesis, we implement a system to correct accent, pronunciation, and intonation in spoken English pronounced by nonnative English speakers. In case of the accent evaluation, energy and pitch information are used to find stressed syllables, and then we extract the segment information for input patterns using a dynamic time warping method to discriminate and evaluate accent position. For the pronunciation evaluation, we utilize the segment information using the same algorithm as in accent evaluation and calculate the spectral distance for each corresponding phoneme between input and reference speech. For the intonation evaluation, we propose nine patterns of pitch slope to model various pitch contours, then we grade test sentences by utilizing accumulated error obtained by the distance measure and estimated slope patterns. Our result shows that 98 % of accent and 71 % of pronunciation evaluation agree with perceptual evaluations. And for the result of the intonation evaluation, our system produces similar grades for four sentences having different intonation patterns to the perceptual evaluations.