中文摘要：

受最近特征线分类器的基本设计思想启发，将最近邻法向平面和空间推广，提出了原点无关最近特征平面和原点无关最近特征空间分类器．与最近特征平面、最近特征空间分类器相比，原点无关最近特征分类器最大的优势就在于其定义的特征子空间不依赖于原点位置，而仅由同模式的若干个特征点决定．这种定义提高了相应的模式相似性度量的合理性与有效性．以人脸识别为例，对多种分类器的性能进行比较．实验结果表明，原点无关最近特征分类器在识别率、稳定性等方面均优于同阶的最近特征分类器．

英文摘要：

With the inspiration of the basic design philosophy of the nearest feature line （NFL） classifier, the geometrical concept of line is generalized to plane and space. And then the paper presents the plane and space versions of nearest neighbor （NN） classifier, namely the origin-independent nearest feature plane （OINFP） classifier and origin-independent nearest feature space （OINFS） classifier. The advantage of OINF classifiers over their counterparts of the same rank, nearest feature plane （NFP） and nearest feature space （NFS） classifiers, lies in their origin-independent feature subspaces definitions because of which their corresponding similarity measures are more reasonable and effective than those of NF classifiers. Several human face recognition experiments have been conducted to compare the performance of various geometrical concept-based classifiers. Experimental results show that OINF classifiers are of higher recognition rate and more stable than NF classifiers.