带有手写干扰的印刷体数字识别

The Recognition of Figures

首先提出黑白特征分离（即黑白特征局部和整体分离提取特征的思想）和一种ＢＰ神经网络的快速收敛算法及训练策略，并在此基础上构造了５种独立的识别模块．应用加权表决方法设计了识别带有手写干扰的印刷体数字及一些特殊字符的综合方法，取得了理想的效果．

A blackwhite feature separate concept(that is extractting blackwhite feature locally and totally) and a BP neural network fast algrithm and trian strategy are proposed. Five independent matching models are further constructed. Then a roubst method, which is used to recognize figures and special symbols interfered by handwritting, is developped and satisfactory results are obtained. 