Abstract

This research aims to explore and implement a personalized speech synthesis model tailored for Korean language learners. Despite significant advancements in general speech synthesis technology, the quality and naturalness of speech synthesis remain challenging for Korean language learners. In this study, we employ deep learning techniques and combine research on facial muscle movements with speech learning to design an innovative framework for personalized speech synthesis. Initially, a substantial amount of speech data from Korean language learners is collected and subjected to preprocessing and annotation. Subsequently, we construct a personalized synthesis model based on deep neural networks to achieve pronunciation correction and fluency improvement for individual learners. The novelty of this research lies in the integration of facial muscle movements with speech learning, leading to optimization in personalized speech synthesis. This innovation holds vital practical implications for enhancing Korean language learners' pronunciation and improving their language communication skills.