Foreign Accent Conversion - Chenxi Liu

This book provides a comprehensive and systematic treatment of Foreign Accent Conversion (FAC), an emerging field at the intersection of speech synthesis, deep learning, and speech signal processing. It begins with an accessible introduction to the foundational concepts of accent and its acoustic characteristics, including temporal, spectral, prosodic, and integrated speech features. This book then presents a structured survey of FAC methods, progressing from traditional approaches—such as voice morphing, articulatory synthesis, and frame-pairing—to state-of-the-art deep learning techniques, including sequence-to-sequence models, TTS-guided methods, and discrete speech token-based approaches. Each method category is explained with its core pipeline, advantages, limitations, and representative studies. This book also covers commonly used datasets and evaluation metrics (both subjective and objective), critically examines the current limitations and challenges facing the field, and outlines promising future research directions. A dedicated chapter provides hands-on resources—including publicly available datasets and code repositories—to help newcomers get started with FAC research. Written for both beginners and experienced researchers, this book serves as an essential reference for anyone interested in understanding, developing, or applying accent conversion technology.