用神经网络方法进行汉字的部件识别

RECOGNITION OF PARTS OF CHINESE CHARACTERS BY NEUNAL NETWORK

用二维图形结构的汉字属性关系图,从二维角度来直接描述汉字具有较强的描述能力.但其中由笔划间的关系图描述的部件通常关系复杂,其匹配问题如果采用传统的匹配方法来解决,其计算量和复杂度是非线性递增的.在此,提出一种神经网络的解决方法.实验结果表明,将神经网络的方法用于汉字的部件识别过程,可以大大地减少计算时间和计算量,且具有一定的容错能力.

The description of Chinese characters can be properly done by a two-dimension. The Attributed Relational Graphs and from a two-dimension graphic structural facet. The calculation and complexity of the matching process increase nonlinearly by the general matching method just because of the complex relation of the branches described by the ARG of strokes. A new method with Neural Network to the branch matching is presented here. The results show that the new method can decrease the calculating time and even h...