知识基联机自然手写汉字识别

A Knowledge Based Agorithm for Online Cursive Chinese Character Recognition

本文研究了一种识别联机自然手写汉字的方法 ,充分利用汉字知识 ,构造联机汉字的知识框架 ,在专门知识引导下识别自然手写汉字。汉字的多种构形规律和书写规律被用于联机汉字的知识表达和建模。输入汉字经过笔划元提取后 ,采用框架描述 ,与知识库中存储的汉字模型进行约束满足的不精确匹配。对国际一级汉字进行测试 ,在对书写次序、笔划数目、是否连笔等很少限制的情况下 ,能获得很高的识别率 ,证明本文方法是有效的。

This paper provides a knowledge guide algorithm for Online Cursive Hand-writing Chinese Character recognition.Many kinds of knowledge that from different standing points are taken into consideration,that include the dynamic writing knowledge and shape forming knowledge.Every character is represented by a framework,and the syntactic and semantic model information are stored in it.Then a inexact constraint satisfaction reasoning algorithm is used to match an input character to the models.Tests on a limit set of GB80-2312 Chinese Character Set demonstrate that the proposed algorithm achieves a high correct rate where with no restrict on the number of strokes,writing orders,and can adapts to large shape variations.