姓名:徐明星

职称:副教授

电话:6279580

邮箱:xumx@tsinghua.edu.cn

个人主页:http://media.cs.tsinghua.edu.cn/~xumingxing/

教育背景

工学学士 (计算机科学与技术), 清华大学, 中国, 北京, 1995;

工学硕士 (计算机应用技术), 清华大学, 中国, 北京, 1999;

工学博士 (计算机应用技术), 清华大学, 中国, 北京, 1999.

研究领域

语音识别, 说话人识别, 语音情感计算

研究概况

语音是人与计算机间最便捷的交互方式,它承载着丰富的信息,如句子的语义信息、说话人的声纹信息、情感状态等。我的研究小组试图以一种统一的框架从语音信号中抽取并识别这些信息,以达到理解语音内容、辨认或确认说话人及其情感状态等目的。我试图探究语音交流过程中的潜在机理,特别是语音的结构化信息。

我的研究小组目前还关注鲁棒说话人识别中的一些问题,如因为信道差异、情感状态变化以及发音方式不同等因素引起的训练与识别的不匹配。我们在这方面提出了基于群集的说话人模型合成方法和情感属性投影算法。在语音情感识别方面,我的研究小组还提出了基于GMM超向量的SVM方法,以及基于ANN的决策融合方法。

学术成果

[1] Wei Wu, Thomas Fang Zheng, Mingxing Xu, Frank Song. A Cohort-based Speaker Model Synthesis for mismatched Channels in Speaker Verification. IEEE Trans. on Audio, Speech and Language Processing, vol. 15, no. 6, pp. 1893-1903, 2007.

[2] Lu Xu, Mingxing Xu, Dali Yang. ANN based decision fusion for speech emotion recognition. Proc.12th Euro. Conf. on Speech Communication and Technology (InterSpeech 2009), Brighton UK, 2009, pp. 2035-2038.

[3] Hao Hu, Mingxing Xu, Wei Wu. GMM supervector based SVM with spectral features for speech emotion recognition. Proc.32nd IEEE Intl.Conf. on Acoustics, Speech, and Signal Processing (ICASSP 2007), Honolulu, Hawaii, USA, 2007, pp. 413-416.

[4] Hao Hu, Mingxing Xu, Wei Wu. Fusion of global statistical and segmental spectral features for speech emotion recognition. Proc. 10th Euro. Conf. on Speech Communication and Technology (InterSpeech 2007), Antwerp, Belgium, 2007, pp. 2269-2272.

[5] Huanjun Bao, Mingxing Xu, Fang Zheng. Emotion attribute projection for speaker recognition on emotional speech. Proc. 10th Euro. Conf. on Speech Communication and Technology (InterSpeech 2007), Antwerp, Belgium, 2007, pp. 758-761.

[6] Wei Wu, Fang Zheng, Mingxing Xu, et al. Study on speaker verification on emotional speech. Proc. 9th Intl. Conf. on Spoken Language Processing (ICSLP 2006), Pittsburgh, Pennsylvania, USA, 2006, pp. 2102-2105.