基于因子分析与稀疏表示的多姿态人脸识别  

Multi-view face recognition based on factor analysis and sparse representation

作　　者：张疆勤[1] 廖海斌[2] 李原[1] 

机构地区：[1]安徽科技学院网络中心,安徽滁州233100 [2]武汉大学电子信息学院,武汉430072

出　　处：《计算机工程与应用》2013年第5期154-159,共6页Computer Engineering and Applications

基　　金：中央高校基本科研业务费专项资金资助(No.20102120103000004);河南省重大科技攻关项目(No.072SGZS38042)

摘　　要：在非可控环境下,人脸识别面临的最大难题之一是姿态变化与遮挡问题。基于稀疏表示的人脸识别方法将测试人脸表示成训练人脸的稀疏线性组合,根据其组合系数的稀疏性进行人脸识别。该方法对人脸的噪声和遮挡变化具有很好的鲁棒性,但对人脸的姿态变化表现力极差,这是因为当人脸具有姿态变化时,同一个人不同姿态情况下很难对应起来,这违背线性组合的前提条件。为了克服稀疏表示方法对人脸姿态变化表现力极差问题,对人脸进行因子分析,分离出人脸姿态因子,得到合成的正面人脸;利用稀疏表示进行人脸分类识别。实验结果表明,该方法对人脸的遮挡和姿态变化具有很好的鲁棒性。Under uncontrolled environment, one of the greatest remaining research challenges in face recognition is to recognize faces across different poses and occlusion. The face recognition method via Sparse Representation （SRC） considers that the test image can be represented as a sparse linear combination of the training images, and further use the combination coefficients for face recognition. This method is robust to face occlusion and noise, but poor performance to face pose varying. The reason is that SRC requires exact alignment between each testing and training image, the variation of pose results in alignment error which is contrary to the prerequisite of the linear combination. In order to overcome of face occlusion and pose variations prob- lem, this paper applies factor analysis to human faces to separate the face pose factors and obtain virtual frontal faces for SRC face recognition.Experimental results demonstrate that the presented algorithm possesses good robustness for the face variation of poses and occlusion.

关 键 词：人脸识别 因子分析 因子分离 稀疏表示 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]