Mispronunciation Detection with Multiple Applications: for Language Learning and Speech Recognition Adaptation and Improvement

This book includes most of Dr. Zhenhao Ge's PhD work in the area of mispronunciation detection, for both language learning and speech recognition adaptation and improvement. In the applications for language learning, this book introduced one type of context adaptive features, that can be used to improve the accuracy of detection mispronunciations. It also described a method based on Principle Component Analysis, to hierarchically detect the parts of mispronunciations for language learners, in an efficient and accurate way. In the applications for speech recognition adaptation and improvement, this book explained how to improve name recognition by learning acceptable pronunciation variations. It also provided practical approaches based on Gaussian Mixture Modals and Linear Discriminative Analysis for accent detection and classification, in order to make speech recognition adaptive to accents and further improve performance.