— About Me

Qing Yao

Hi, I'm Qing Yao, a Ph.D. student at Jiangsu University. My research focuses on generative speech enhancement, aiming to achieve high-fidelity speech reconstruction through efficient generative modeling.

I explore two complementary directions: continuous diffusion models, particularly Schrödinger bridges, operating in the complex-valued time–frequency domain, and discrete masked diffusion models operating on acoustic tokens extracted from neural audio codecs.

Motivated by the ill-posed nature of speech enhancement, I seek to develop generative approaches that balance reconstruction fidelity, perceptual naturalness, and computational efficiency, while improving generalization across diverse acoustic conditions.