手写印刷体汉字识别方法2-D EAG  

HAND PRINTED CHINESE CHARACTER RECOGNITION METHOD 2-D EAG

作　　者：赵明[1] 

机构地区：[1]中国科学院软件研究所

出　　处：《软件学报》1991年第3期11-19,共9页Journal of Software

基　　金：国家自然科学基金

摘　　要：本文介绍手写印刷体汉字识别方法2-D EAG。主要的工作为:提出了二维扩展属性文法模式识别方法,可实现自底向上归约和自顶向下推导双向信息传递和控制;提出了一种直接利用二维信息进行识别的方式,可避免特征线性化造成的信息丢失;提出了多义文法、共生文法和结构推断三种在文法之间建立联系,利用结构类比区分极相似字的 识别算法;提出了多冗余归约机制,双边缘弹性跟踪笔段抽取算法,多冗余有引导部件抽取算法。用2-D EAG方法对100字种实际手写汉字进行了识别实验,在识别大畸变汉字(包括部分连笔字)方面得到了很好的结果。This paper introduces 2-D EAG method for hand printed Chinese character recognition. The main works are: Propose Two-Dimensional Extended Attribute Grammar (2-D EAG for short) method, in which both bottom-up reduction And top-down deduction are realized for two-way information transmitting and control; Propose a recognition scheme utilizing directly two dimensional information, to avoid information losses caused by linearization of features; Provide polysemous grammars, coexisting grammars and structure inferrings which build special linkages among grammars and distinguish among similar samples by structural analogy; Besides, realize also redundant reduction scheme, stroke segment drawing algorithm with elastic tracing of contour lines, and redundant guided component drawing algorithm. The recognition experiment with 100 categories of hand printed Chinese characters is conducted and very good result is achieved in overcoming large scale structural distortions (including some cursive writings).

关 键 词：手写印刷体 汉字识别 计算机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]