基于MW(2D)~2 PCA的单训练样本人脸识别

传统的人脸识别方法在单训练样本条件下性能会急剧下降,因此,研究出适合于单样本情况下的识别算法是人脸识别问题面临的巨大挑战.针对两个方向的二维主成分分析((2D)~2PCA)算法进行改进,文中提出将加权和分块与(2D)~2PCA相结合的方法称为分块加权(2D)~2PCA,以便更有效地提取人脸的局部特征.同时把模糊理论引入分类决策,应用于单训练样本人脸识别问题.在ORL人脸库以及部分CAS-PEAL人脸库中的实验结果表明,文中方法能取得较好的识别效果.

MW(2D)~2 PCA Based Face Recognition with Single Training Sample

Traditional methods get low recognition accuracy in the condition of only one training sample,therefore,it is a great challenge for face recognition.In this paper,a two-directional two-dimensional principal component analysis((2D)~2PCA)is developed to solve this problem.An improved arithmetic combining weight and block called modular weighted(2D)~2PCA is proposed for efficient local feature extraction.Besides,the fuzzy theory is introduced to classify the single sample face recognition.Experimental results on ORL and a subset of CAS-PEAL face databases show that the presented method achieves a high recognition accuracy.