Yi Xu

Yi Xu is a Professor of Speech Sciences at the Division of Psychology and Language Sciences, University College London (UCL), where he investigates the fundamental mechanisms of speech production and perception in connected discourse. His research centers on how multiple layers of communicative meaning are encoded through articulation, with pioneering contributions to speech prosody modeling including the Target Approximation (TA) model and Parallel Encoding and Target Approximation (PENTA) model. He also developed the Bio-informational Dimensions (BID) theory for vocal expression of emotions and maintains active international collaborations on cross-linguistic prosody phenomena.

Professor Xu's research spans speech prosody, computational phonetics, emotional speech, and coarticulation, focusing on dynamic articulatory mechanisms in tonal languages like Mandarin. His work integrates theoretical modeling, empirical validation, and tool development to address core questions about meaning encoding in speech. Key themes include the relationship between articulatory constraints and perceptual outcomes, neural processes in speech acquisition, and evolutionary perspectives on emotional expression.

Recent publications (2023-2024) demonstrate strong interdisciplinary convergence across linguistics, computer science, and cognitive science, with emphasis on Mandarin tone-syllable interactions, deep learning applications in phonetic analysis, and cross-linguistic timing patterns. His work consistently bridges theoretical frameworks with practical computational implementations, advancing understanding of speech as a multi-layered communication system.

He has created influential research tools including qTAtrainer for automatic intonation analysis based on the TA model, PENTAtrainer for function-loaded pitch target extraction, and FormantPro for large-scale formant movement analysis. These Praat-based tools are widely adopted in speech science for prosody research and have enabled large-scale empirical studies of connected speech phenomena.