唇读研究进展综述  

Review of lip-reading research development

作　　者：张泽梁[1] 杨成佳[2] 宋绍成[1] 

机构地区：[1]北华大学信息技术与传媒学院,吉林吉林132011 [2]吉林大学计算机科学与技术学院,吉林长春130000

出　　处：《计算机工程与设计》2014年第6期2135-2141,共7页Computer Engineering and Design

基　　金：国家科技支撑计划子课题基金项目(2011BAK07B03-9)

摘　　要：将计算机唇读技术和语音识别技术进行融合,提高语音识别效果的方法,引起了一些研究者的关注,并已取得了长足的进步,但仍有许多难题需要突破。为了引起更多研究者对此研究领域的兴趣和关注,参与到唇读技术的研究,共同推进该领域的发展,就目前唇读技术的发展现状做了详细的介绍。总结一些主要的传统方法及相关方面的新技术,主要关注点为视觉特征提取方法、识别技术和音视频的信息融合算法。To improve speech recognition results,the approach fusing lip-reading computer technology and speech recognition technology was introduced.This method attracted the attention of some researchers,and considerable progress was made,but there were still many problems to be brokethrough.To attract more researchers＇ interests and concerns in this area of research and hoped that they could participate in the study of lip-reading technology to jointly promote the development of the field,a detailed introduction on the current status of lip-reading technology was given,some of the main relevant aspects of the traditional methods and new technologies were described.It focused on the visual feature extraction,recognition technology and speech visual fusion algorithr.

关 键 词：唇读 视觉特征 特征提取 隐马尔可夫模型 信息融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]