计算机唇读研究进展  

Research Advances in Computer Lip-Reading

作　　者：王晓平[1] 郝玉峰 付德刚[1] 袁春伟[1] 

机构地区：[1]东南大学生物电子学国家重点实验室,南京210096 [2]北京捷通华声语音技术有限公司,北京100094

出　　处：《数据采集与处理》2007年第3期353-359,共7页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金(60121101)资助项目

摘　　要：计算机唇读是利用计算机对说话者的唇动等视觉语音信息进行分析以识别出其所说内容的过程,并可与听觉语音信息相融合以进一步提高计算机的识别率,从而使人机交互更加自然。本文从计算机唇读系统的各环节入手综述了该领域的研究进展,并讨论了现有诸方法的优缺点,最后提出了有待进一步研究的问题。As a hotspot in the field of human-computer interaction, computer lip-reading aims at recognizing what human says by analyzing visual speech information, such as lip movement. It can be further integrated with audio speech information to improve recognition accuracy for more convenient human-computer interaction. This paper gives a survey of lip-reading approaches and discusses their benefits and drawbacks. Finally, several key issues to be researched in the field are pointed out.

关 键 词：唇读 定位 特征抽取 信息融合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]