Academic Lecture—Speech enhancement method combining binaural auditory perception

• 发布时间：[2025-04-11] 来源：[学院] 点击量：[1463]

Presentation Title: Speech Enhancement Methods Incorporating Binaural Perception

Presentation Time: 2025/4/11 15:00 China Standard Time - Beijing

Venue: Conference Room 401, School of Electronic Information, Wuhan University (Information Science Department, West Building)

Speaker: Dr. Pan Ningning

Inviter: Professor Huang Gongping

Abstract:

In intelligent interaction and voice communication systems, speech signals received by microphones are inevitably contaminated by noise, which in turn reduces intelligibility. In complex acoustic environments such as strong noise or multiple sources, recovering the desired speech signal from the noisy signal and improving intelligibility is a challenging problem. Research in psychoacoustics shows that appropriate binaural presentation in the auditory perceptual space can significantly improve speech intelligibility. This talk will share how to design binaural speech enhancement methods that incorporate the characteristics of human binaural hearing. Specifically, we will introduce a deep learning-based spatial sound source orientation method and a linear binaural speech enhancement method under interaural coherence constraints.

About the Speaker:

Ningning Pan received her bachelor's, master's, and doctoral degrees from Northwestern Polytechnical University in 2014, 2017, and 2023, respectively. From 2018 to 2020, she was jointly trained at Columbia University and Georgia Institute of Technology. She joined Southwestern University of Finance and Economics in 2023 and is currently a lecturer and master's supervisor at the School of Computer Science and Artificial Intelligence. Her research focuses on speech enhancement, binaural hearing, and music generation.