基于加权不相关鉴别分析的人脸识别  

Face recognition based on uncorrelated weighted discriminant analysis

作　　者：梁毅雄[1] 龚卫国[1] 潘英俊[1] 李伟红[1] 

机构地区：[1]重庆大学光电技术及系统教育部重点实验室,重庆400044

出　　处：《计算机应用》2005年第8期1764-1766,共3页journal of Computer Applications

基　　金：教育部科学技术重点资助项目(02057);教育部春晖计划资助项目(2003589)

摘　　要：提出了一种基于加权不相关鉴别分析的人脸识别方法。该方法引入了一种新的权函数对Fisher准则加权,以改善样本在低维线性空间中的可分性;然后,以给出的加权Fisher准则为目标函数,在共轭正交的约束下求解其最佳投影方向,从而保证所提取的最佳鉴别特征之间的统计不相关性。实验结果表明,与经典的特征脸方法和Fisher脸方法相比,该方法对光照变化、表情变化以及时间变化等不敏感,具有更好的鲁棒性。A novel method based on weighted uncorrelated discriminant analysis for face recognition was proposed. First, the Fisher criterion was redefined by introducing a weighting of the contributions of individual class pairs to the overall criterion. Then, the weighted Fisher criterion was optimized under the conjugated orthogonal constrain, which can guarantee the derived projection directions were statistically uncorrelated. Experiments were carried out to compare the proposed method with classical Eigenfaces method and other LDA-based methods such as Fisherfaces and ULDA. The experimental results on the AR face database show the effectiveness of the proposed algorithm and its insensitivity to the variants of face expression, illumination and sessions.

关 键 词：线性鉴别分析 加权Fisher准则 人脸识别 统计不相关 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]