计算机应用研究2017，Vol.34Issue(4)：1221-1225,1230,6.DOI:10.3969/j.issn.1001-3695.2017.04.060
高安全性人脸识别系统中的唇语识别算法研究
Lip-reading algorithm in face recognition systems for high security
摘要
Abstract
Aiming at defending against malicious attacks on face recognition systems using photos and videos,this paper proposed a high security face recognition system,which combined face recognition and lip-reading based liveness detection.Firstly,due to the lack of Chinese lip-reading data,it established two large Chinese lip-reading datasets named CNLIP1 and CNLIP2.Secondly,in order to extract temporal information from the data,it used an extensive version of the stacked convolutional independent subspace analysis algorithm to learn features from video data.Finally,it proposed to use transfer learning algorithm to train a speaker-dependent lip-reading model.Experiments show that the proposed feature gives preferable performance,the model can meet the needs of the liveness detection and the constructed high security face recognition system has good anti-attack effect.关键词
高安全性/人脸识别/活体检测/唇语识别/迁移学习Key words
high security/face recognition/liveness detection/lip-reading/transfer learning分类
信息技术与安全科学引用本文复制引用
任玉强,田国栋,周祥东,吕江靖,周曦..高安全性人脸识别系统中的唇语识别算法研究[J].计算机应用研究,2017,34(4):1221-1225,1230,6.基金项目
中国科学院战略性先导科技专项基金资助项目(XDA06040103) （XDA06040103）
国家自然科学基金资助项目(61472386) （61472386）
国家自然科学基金青年基金资助项目(61502444) （61502444）
