An Intelligent System for English Pronunciation Correction

摘要

English pronunciation correction highly relies on the English teaching quality, therefore, in this paper, we propose a novel intelligent system for English pronunciation correction based on both Convolutional neural networks (CNN) and voiceprint proof. CNN refers to a multi-layer neural network, in which each layer contains multiple 2D planes and each plane is made up of a lot of independent neurons. We read the voice signal of speakers in advance, and then handle the original voices based on the CNN model. Then, the MFCC features are extracted and input to the CNN and voiceprint recognition is achieved by analyzing the output of CNN. Afterwards, English pronunciations are corrected through comparing the voiceprint of the target user to the standard pronunciation. Finally, experimental results demonstrate that the proposed method is able to achieve high voiceprint recognition rate for various experiment settings.