核最近邻凸包分类算法

Kernel Nearest Neighbor Convex Hull Classification Algorithm

为了增强最近邻凸包分类器的非线性分类能力,提出了基于核函数方法的最近邻凸包分类算法。该算法首先利用核函数方法将输入空间映射到高维特征空间,然后在高维特征空间采用最近邻凸包分类器对样本进行分类。最近邻凸包分类器是一类以测试点到各类别凸包的距离为相似性度量,并按最近邻原则归类的分类算法。人脸识别实验结果证实,这种核函数方法与最近邻凸包分类算法的融合是可行的和有效的。

A novel pattern classification algorithm based on the kernel method named kernel nearest neighbor convex hull （ KNNCH ） algorithm is presented in this paper. First, the data from the input space are projected into a higher dimensional feature space by replacing the inner product with an appropriately chosen kernel function. Then, the nearest neighbor convex hull（NNCH） classifier is constructed for classification in the higher dimension feature space. In NNCH classifier, the distance between a test sample and a convex hull of training samples of a class is taken as the similarity measure for classification. According to the nearest neighbor rule, a test sample will be classified to the class of the nearest convex hull. The experiments on face recognition show good performance of the combination of kernel method and NNCH classifier.