摘要： 受支持向量机的几何解释和最近点问题启发，提出一种新型的模式分类算法——核仿射子空间最近点分类算法。该算法在核空间中，将支持向量机几何模型中的最近点搜索区域由2类训练特征集凸包推广到2类特征样本各自生成的仿射子空间，以仿射子空间作为特征样本分布的粗略估计，通过仿射子空间中的最近的2个点构造平分仿射子空间间隔的最优分类超平面。该算法在ORL人脸识别数据库上的比较实验中取得了较好的识别效果。

关键词: 模式分类, 核函数, 支持向量机, 核仿射子空间最近点

Abstract: A novel pattern recognition algorithm called Kernel Affine Subspace Nearest Points(KASNP) classification is presented. Inspired by the geometrical explanation of Support Vector Machine(SVM) that the optimal separating plane bisects the closest points within two class convex hulls, KASNP algorithm expands the searching areas of the closest points from the convex hulls to their corresponding class affine subspaces in kernel space. The affine subspaces are taken as the rough estimations of the class feature sample distributions, and their closest points are found. The hyperplane to separate the affine subspaces with the maximal margins is constructed, which is the perpendicular bisector of the line segment joining the two closest points. The test experiments compared with the Nearest Neighbor(1-NN) classifier and SVM on the ORL face recognition database show good performance of this algorithm.

Key words: pattern classification, kernel function, Support Vector Machine(SVM), Kernel Affine Subspace Nearest Points(KASNP)

中图分类号: 

• TP391