Academic Lecture—Research Progress on Large-Scale Models for Speech Generation

Author：Administrator Source：website Time：2025-01-18 12:00:00

Presentation Time: 11:00 AM, Saturday, January 18, 2025

Venue: Conference Room 501, School of Electronic Information, Wuhan University (Information Science Department, West Building)

Presentation Title: Research Progress on Unified Models for Speech Processing and Generation

Speaker: Dr. Wu Zhengzhi

Inviter: Professor Huang Gongping

Abstract:

With the widespread application of speech generation technology, systems and products such as human-computer interaction and AIGC are increasingly demanding on the expressiveness, security, and reliability of generated speech. As we all know, speech signals contain both linguistic and paralinguistic information. Therefore, generating expressive, secure, and reliable speech requires jointly understanding and accurately modeling both linguistic and paralinguistic information. However, research in this area is still in its infancy. This presentation will share research progress on large-scale speech generation models that are capable of zero-shot learning, are highly expressive, and are secure and reliable.

About the Speaker:

Wu Zhizheng is an associate professor and doctoral supervisor at the Chinese University of Hong Kong, Shenzhen, and a national-level young talent. He has been repeatedly selected as one of Stanford University's "Top 2% Global Scientists." He holds a doctorate from Nanyang Technological University and has held academic research and technical leadership positions at institutions such as Meta (formerly Facebook), Apple, the University of Edinburgh, and Microsoft Research Asia. He initiated the Merlin and Amphion open-source systems and the Emilia open-source dataset, which have been adopted by over 300 institutions and have repeatedly topped GitHub trending lists. He has also led international evaluations of speech forgery detection, speech synthesis, and voice conversion, winning numerous best paper awards. He serves on the editorial boards of journals such as IEEE/ACM TASLP and SPL, and chaired the SLT2024 conference.