GENERAL INFORMATION
| 2008.12 - | Associate professor. |
| 2014.08 - 2015.07 | Visiting Scholar, Beckman Institute, University of Illinois at Urbana-Champaign. [link] |
| 2003.08 - 2008.12 | Assistant professor, Department of Electronic Engineering, Tsinghua University, Beijing, China. |
| 1998.09 - 2003.07 | Ph.D. in electronic engineering from Tsinghua University, Beijing, China. |
| 1994.09 - 1998.07 | B.S. with the highest honor in electronic engineering from Shanghai Jiaotong University, Shanghai, China. |
RESEARCH INTERESTS
- Speech Processing (speech recognition and understanding, speaker recognition, natural language processing)
- Statistical Machine Intelligence (particularly with graphical models)
I'm broadly interested in Dealing with Uncertainty for Intelligent Systems. The two fundamental problems are first to construct probabilistic models of the studied phenomenon using human knowledge and machine learning algorithms, and second to find efficient ways of implementing probabilistic inference with those models. Graphical models have emerged as a general framework for describing and applying probabilistic models, while the core is still probability theory and mathematical statistics.
Current interests are in Speech Processing, which is of great value for scientific research, commercial application and national defense. 100,000 years ago, humans started to speak so they could free up their hands. Is it going to happen again? Getting rid of the keyboard and talking to machines (desktop PCs, hand-held devices, and remote servers) is not a dream in the near future, I believe. Moreover, efficient searching, mining, and browsering the large amounts of digital media, in which speech is often an indispensable component, also needs speech signal and information processing techniques.
人类智能本质是否在于能有效处理不确定性，这个问题还有待进一步研究论证。但处理不确定性是智能的一种重要表现——这个认识已经得到众多研究者的肯定和践行。我广泛感兴趣于智能系统中的不确定性处理——统计机器智能。这涉及两个基本问题，一是借助人类知识和机器学习算法，建立所研究对象的概率模型，二是基于建立的模型，寻找有效概率推理的方法。图模型已逐渐成为描述和应用概率模型的一般框架，其核心是概率论和数理统计。
目前研究兴趣在语音信号与信息处理。语音处理是机器智能的一个重要切入点，对商业和国防也有着非常重要的实用价值。十万年以前，人学会用言语进行交流，释放了双手，从猿进化到人，这是人类文明的重要一步。当今人类进入计算机广泛使用的信息时代，如何能让人的双手不再被键盘鼠标束缚，人能与机器进行自然的言语人机交互？（这里机器包括桌面电脑，手持设备，也可以是网络端服务器。）我相信，这在不久的将来不再是一个梦想，将带来人类文明的又一大进步。此外，音频是多媒体的不可缺少的一部分。对日益增多的音视频内容进行有效地搜索、挖掘和浏览，也非常需要语音信号与信息处理技术。
RESEARCH PROJECTS
TEACHING
- Theory and Applications of Probabilistic Graphical Models, since 2004 Fall.
- Probability and Stochastic Processes (1), since 2012 Spring.
- Probability and Stochastic Processes (2), since 2012 Fall.
- Mathematics and engineering applications, 2005 Spring - 2011 Spring.
PROFESSIONAL ACTIVITIES
- IEEE Senior Member.
- 中国计算机协会（CCF）语音对话与听觉专委会委员.
- 全国人机语音通讯会议常设机构委员会委员.
HONORS
- 2018年，指导ISCSLP (中文言语处理国际会议) Best Student Paper Award。
- 2018年，指导清华大学优秀博士论文。
- 2012年，清华大学第五届青年教师教学大赛一等奖。
- 2010年，清华大学优秀班（级）主任一等奖。
- 2005年，全国人机语音通信会议优秀论文。
- 2004年，特定内容音频识别系统(集成语音关键词识别、说话人识别、音频指纹)通过国家计算机网络与信息安全管理中心的评测。
- 2003年，赢得2003年度国家863语音识别评测－音节识别任务。
- 2002年，获清华大学电子系博士生学术论坛最佳论文奖。