SPATIAL AUDIO · MACHINE LEARNING
Xingyu Chen
PhD Candidate / University of Technology Sydney
I study how machines can reconstruct and generate sound in ways that are grounded in the physical world and meaningful to human listeners.
Sound is shaped by sources, environments, and listeners. A recording captures only a partial observation of their interaction. My research asks how to represent this underlying structure so that we can infer beyond what is observed and generate sound with meaningful control. I am interested in what should be constrained by acoustics, what should be learned from data, and what must be preserved for human perception.