Aug.1995 - Jul.1999  M.E. School of Computer Science, Tsinghua University, Major in Computer Application Technology

Sep.1990 - Jul.1995  B.E. School of Computer Science, Tsinghua University, Major in Computer Science and Technology

Jul.2020 - Present Associate Researcher, Beijing National Research Center for Information Science and Technology, Tsinghua.

Jun.2018 - Jun.2023  Deputy Director, Organization Department of the Party Committee, Thua University,Associate Researcher

Dec.2003-Jun.2018  Associate Professor、 Associate Researcher, Department of Computer Science, Tsinghua University; Served as the Party Group Leader for Graduate Students, Deputy of the Party Committee, and Chairman of the Trade Union.
Aug.1999 - Dec.2003  Lecturer, Department of Computer Science, Tsinghua University

（1）Member of the Speech Dialogue and Hearing Special Committee of the Chinese Computer Society

（2）Member of Emotional Intelligence Special Committee of Chinese Artificial Intelligence Society

（3）Member of the Music Psychology Special Committee of the Chinese Psychological Society

（4） Secretary General of the Financial Technology Special Committee of the China Electronic Information Industry Federation

（1）Computer application technology
（2）Artificial intelligence
（3） Human-computer interaction

The main research directions include voice recognition, speaker recognition, voice and music emotion recognition, cross media emotion computing, Internet discourse emotion recognition, etc..

A multi-scale fusion emotion detection method is proposed to address the dynamic emotional changes in dialogue speech. A decision level fusion speech emotion recognition framework with multiple features and classifiers is proposed for speech emotion recognition in real environments. From different perspectives of temporal and spatial multiscale, this study explores the extraction and modeling of structural information in audio signal time series. Based on a hierarchical cognitive mechanism, a dynamic dimension emotion recognition algorithm is proposed that integrates multi-scale contextual information using attention mechanism. Based on the Long Short Term Memory (LSTM) model, we integrate deep neural networks and attention mechanisms to integrate the fusion process of multi-scale information with the context modeling process at different scales into a single algorithm framework. We jointly optimize the modeling process of contextual and structural information to achieve dynamic sentiment modeling and recognition across dimensions.

In terms of cross media sentiment computing, a cross media unsupervised feature learning algorithm based on heterogeneous entropy is proposed, focusing on the correlation between cognitive representation and semantic description of cross media data. This algorithm integrates cross media data such as text, images, and network behavior to predict the emotional style (personality) of social users.

In terms of robust speaker recognition research, we focus on the mismatch between training and recognition caused by channel differences, emotional state changes, and different pronunciation methods. We have developed a speaker recognition speech library with multiple pronunciation methods, proposed a cluster based speaker model synthesis method, and proposed an emotional attribute projection algorithm to improve the robustness of speaker recognition to emotional changes.

（1） Second Prize of Beijing Science and Technology Award（2023）

（2） Second Prize of Chinese Industry University Research Cooperation Innovation and Promotion Award（2022）

（3） First Prize of Chinese Institute of Electronics Technology Invention（2021）
（4）Outstanding Party Building and Ideological and Political Worker of Tsinghua University （2017）

（5）Second Prize of Excellent Scientific Research Achievement Award (Science and Technology) for Higher Education Institutions （2016，ranked 8/19)

（6）Second Prize for Teaching Achievements at Tsinghua University （2014）

（7）Tsinghua University Student Laboratory Construction Guidance Award（2014）

（8）Tsinghua University Qingyun Candlelight - Students' Favorite Teacher（2011）

（9） Tsinghua University Lin Feng Counselor Award（2006）

（10）Tsinghua University Young Teacher Teaching Excellence Award（2005）

（11）First Prize of Education and Teaching Achievements (Higher Education) in Beijing(2005，ranked 3/5)

[1] A sentiment representation method based on Gaussian mixture model hypervector and a decision fusion method based on ANN were proposed, and a patent was granted for the results. Won the championship in the audio emotion recognition subtask of the International Audio Video Emotion Recognition Competition (MEC 2017) organized by ACII Asia 2018.

[2] A multi time scale regression analysis algorithm based on bidirectional LSTM and a multi sensory spatial scale regression analysis algorithm based on SVR were proposed. Ranked first in the music emotion recognition evaluation task organized by MediaEval in 2015.

[3] A new recording replay detection algorithm framework was proposed, which ranked first in the anti recording attack challenge task in ASV Spoof 2019.

[4] The achievements related to speech emotion recognition have been applied in the "Unsupervised Identity Authentication System Based on Dynamic Password Speech" to detect users' true intentions. The evaluation opinion of the scientific and technological achievement appraisal organized by the China Electronics Society believes that the overall technology of the identity authentication system has reached the international leading level.

（1） Xinxing Li  (Outstanding Master of Engineering graduate from Tsinghua University, 2017)

（2） Xiaotong Zhang (Outstanding Engineering Master's graduate from the Department of Computer Science, 2019)