基于动态贝叶斯网络的音视频双模态说话人识别-humancomputer.pdfVIP

计算机研究与发展 ISSN N 111777TP Journal of Computer Research and Development 43 (3) : 470～475 , 2006 基于动态贝叶斯网络的音视频双模态说话人识别 吴志勇 　　　蔡莲红 (清华大学计算机科学与技术系普适计算教育部重点实验室 　北京 　100084) (wuzy99 @mailstsinghuaeducn) AudioVisual Bimodal Speaker Identif ication Using Dynamic Bayesian Net works Wu Zhiyong and Cai Lianhong ( Key L aboratory of Pervasive Computin g , Minist ry of Education , Depart ment of Computer Science and Technology , Tsinghua ) U niversity , Beijin g 100084 Abstract 　Studied in this paper is the use of dynamic Bayesian networks (DBNs) for the task of text prompt audiovisual bimodal speaker identification The task is to determine the identity of a speaker from a temporal sequence of audio and visual observations obtained from the acoustic speech and the shape of the mouth respectively According to the hierarchical structure of audiovisual bimodal modeling , a new DBN is constructed to describe the natural audio and visual state asynchrony as well as their conditional dependency over time The experimental results show that the dynamic Bayesian network is a powerful and flexible methodology for representing and modeling the audiovisual correlations and the proposed DBN can improve the accuracy of audioonly speaker identification at all levels of acoustic signaltonoise ratio ( SNR) from 0 to 30dB Key words 　biometrics ; speaker identification ; audiovisual bimodal modeling ; fusion ; dynamic Bayesian network (DBN) 摘 　要 　动态贝叶斯网络在描述具有多个通道的复杂随机过程方面具有优异的性能 基于动态贝叶斯 网络进行音视频双模态说话人识别的工作 分析了音视频联合建模的层级结构 ,利用动态贝叶斯网络 对不同层级的音视频关联关系建立模型 ,并基于该模型进行音视频说话人识别的实验 通过对不同层 级的建模过程及说话人识别实验的结果进行分析 ,结果表明 ,动态贝叶斯网络为描述音视频间的时序相 关性和特征相关性提供了有效的建模方法 ,在不同语音信噪比的情况下均能提高说话人识别的性能 关键词 　生物识别 ;说话人识别 ;音视频联合建模 ;融合 ;动态贝叶斯网络 中图法分类号 　TP