模糊多分类器组合

本文提出了一种多分类器的模糊组合方法,它利用了参与组合的分类器提供的两类信息:(1)在度量层次上,对未知模式的分类信息,(2)在符号层次上,训练样本的错分类分布信息.对参与组合的分类器提供的这两类信息进行模糊集成,组合分类器输出未知模式来自各类别的可能性度量.用该方法对手写体汉字作分类识别,实验结果显示,较之其它几种方法,它有更高的可靠性.

Fuzzy Combination of Multiple Classifiers

This paper presents a fuzzy combination method of multiple classifiers, which uses two kinds of information provided by the applied classifiers, one is for the classification of the unknown patterns on the measurement level, and the other is for the distribution of the wrong classification of the training samples on the abstract level. Those two kinds of information are integrated fuzzily to form the combined classifier outputting the probability measurement of the unknown patterns from all classes. This method is applied to classify and recognize the handwriting Chinese characters, and the experimental results show that this method is superior to the other method on the higher reliability.