Academic Lecture—Speech separation and generalization

• 发布时间：[2025-04-27] 来源：[学院] 点击量：[1467]

Presentation Time: 9:00 AM, Sunday, April 27, 2025

Venue: Conference Room B621, School of Electronic Information, Wuhan University (Yu Gang & Song Xiao Building)

Presentation Title: Speech Separation and Generalization

Presentator: Associate Professor Wang Zhongqiu

Inviter: Professor Huang Gongping

Abstract:

Enabling computers to clearly hear human speech is a critical need in fields such as artificial intelligence and human-computer speech interaction. In noisy environments with multiple sound sources, microphones typically record a mixed signal consisting not only of the target speaker's speech but also interference signals such as ambient noise, reverberation, and human voice interference. These interference signals can severely degrade the quality of speech interaction, speech recognition performance, and the ability of hearing-impaired individuals to perceive the target sound source. Focusing on the challenge of accurately separating the target speech from mixed signals, this presentation will highlight the presenter's research progress in deep learning-based speech separation techniques, covering supervised, unsupervised, weakly supervised, and semi-supervised speech separation, single-channel and multi-channel speech separation, target speaker extraction, and how to improve their generalization capabilities on real-world data.

About the Speaker:

Dr. Wang Zhongqiu, a National Young Talent, is an Associate Professor in the Department of Computer Science and Engineering at the Southern University of Science and Technology. He previously held visiting positions at Mitsubishi Electric Research Institute in the United States and a postdoctoral fellow at the Language Technologies Institute at Carnegie Mellon University. Dr. Wang's research focuses on the perception, understanding, and generation of acoustic signals in artificial intelligence and computational hearing. To date, he has published over 60 papers in leading journals and conferences in the fields of artificial intelligence, speech, and audio signal processing. He was awarded the Best Student Paper Award at ICASSP 2018, the flagship conference on signal processing. For more information, please visit https://zqwang7.github.io/.