模式识别中K-L特征选择的推广及综合距离特征选择的新算法

THE GENERALIZATION OF K-L METHOD IN PATTERN RECOGNITION AND A NEW ALGORITHM FOR COMPOUND DISTANCE FEATURE EXTRACTION

本文主要讨论用欧氏距离作为判据进行特征选择的问题。首先对经常使用的类内、类间距离特征选择方法作了一个统一的叙述,其次着重讨论了综合距离特征选择问题。对于这一问题,本文给出了一个有效的迭代算法,并将其运用于实例进行计算,得到了满意的结果。

In this paper,we focus our attention on the subject of feature extractionbased on the euclidian distance criteria.At first,we give a general review onintraset and interset distance feature extraction,and lay emphasis on compounddistance feature extraction.Following the discussion,an effective iterative algorithmis found,and utilized for solving a real problem with satisfactory results.