一种新的广义最近邻方法研究

An Extended Nearest Neighbor Method Based on Bionic Pattern Recognition

本文在仿生模式识别理论的指导下 ,推广了传统的最邻近方法 ,提出一种新的广义最近邻方法 .该方法利用同类样本间的连续性规律 ,增加了样本点个数 .再通过由多个超球体的并所形成的几何形体 ,覆盖可能的样本点 .将这一方法用于人脸识别仿真实验 ,实验结果表明该方法是有效可行的 .

With the direction of the theory of Bionic Pattern Recognition,this paper puts toward an extended nearest Neighbor method.It makes use of the continuity rule between the samples which belong to the same pattern to increase the number of samples.We cover the possible samples by means of the combination of the hyperspheres.This method is proved effective by simulation result of the face recognition experiment.