Recent Advancement in Accent Conversion Using Deep Learning Techniques: A Comprehensive Review

Accent conversion aims to synthesis a new voice that contains the voice quality of the source speaker but accent of the target speaker. Accent conversion involves multiple speech processing techniques to complete the task like speech analysis, speech feature extraction, train the model using extracted features and resynthesize speech from the converted features. With the latest advancement of technology in speech processing area we can now synthesis human like speech with high similarity of the voice quality with source speaker. In this paper, we provide a comprehensive overview of the accent conversion techniques and their performance evaluation methods based on deep learning and discuss their limitations and promises. This is the first accent conversion literature review as per our best knowledge. In this paper we also provide the details of the available resources for accent conversion research. This study gives a detailed summery of the different implementation of the deep learning methods for accent conversion.