Chunyang Xu is a Lecturer at the Center for Auditory and Language Intelligence, Soochow University. He received his Ph.D. from University College London and previously worked as a Senior Research Engineer at Huawei Technologies. His main research interests include psychoacoustics and auditory perception, sound field reproduction and control, personalized spatial audio, and sound quality design for intelligent audio terminals.
Sound control and reproduction typically take mean-square error or residual acoustic energy minimization as the optimization objective. However, numerical optimality at the signal level does not necessarily correspond to perceptual optimality in real acoustic environments. This talk discusses this discrepancy across four levels: optimization objective, signal-level design, acoustic implementation, and perceptual evaluation. At the signal level, we analyze the relationship between the spectral distribution of the error and auditory quality, and the design trade-off between per-band accuracy and overall error under the constraint of finite-length real-coefficient causal filters. At the acoustic implementation level, we examine the characteristics and variations of real loudspeakers, microphones, and propagation spaces, discuss the problems encountered when translating signal-level optimization results into actual auditory experience, and present ear-canal sound control and reproduction as a key case study. Finally, we look ahead to the potential of audio and multimodal large models in perceptual evaluation, exploring evaluation approaches that combine sound characteristics, scene context, and natural language descriptions. Through typical phenomena and cases, this talk aims to connect signal optimization, physical acoustic constraints, and auditory perception, providing a more complete research perspective for the design and evaluation of sound control and reproduction.