Xmart青年论坛由上海交通大学跨媒体语言智能实验室（SJTU X-LANCE lab）创办，中国计算机学会语音对话专委会主办，语音之家协办，旨在邀请国内外优秀的青年学者分享其最新科研工作和成果，促进多元且深入的交流与合作。Xmart学生论坛作为其中一个系列，致力于邀请国内外知名高校有成体系工作的研究生，主要通过线上分享的方式，系统地介绍其科研成果和心得，为青年学生打造一个学术探讨，思维碰撞和多学科交叉融合的平台。
Xmart•学生论坛丨丁翰林：用于通用音频理解的可扩展方法
形式：线上
时间：9月3日(周三) 14:00 ~ 16:00
报告嘉宾
Heinrich Dinkel (丁翰林) is an Algorithm Engineer at Xiaomi. Before Xiaomi, he obtained his PhD degree from the SpeechLab (now X-Lance) at Shanghai Jiao Tong University, supervised by Kai Yu and Mengyue Wu. His research interests include scalable methods for general audio and multimodal understanding. He has published over 30 papers in international conferences such as ICASSP, Interspeech, and ACM-MM. He is a reviewer for Interspeech, ICASSP, EUSIPCO, DCASE, TASLP, and Speech Communications.
报告摘要
Recent years have seen a significant increase in demand for more versatile audio understanding solutions, moving beyond traditional, task-specific approaches like automatic speech recognition (ASR). The focus is shifting towards general audio understanding, which aims to create a single, unified model capable of handling a wide range of audio tasks. Specifically, we will discuss key design choices for general audio encoders and the associated challenges that arise with data and parameter scale. Further, the talk will introduce encoder-decoder training for general audio understanding, focusing on the integration of a Large Language Model (LLM) as a decoder.
参加方式
①
直播将通过语音之家微信视频号进行直播
手机端、PC端可同步观看
👇👇👇
②
腾讯会议参加
会议号：886-783-088
