基于模糊方向线素特征的手写体汉字识别  

Handwritten Chinese characters recognizing based on fuzzy directional line element feature

作　　者：马少平[1,2] 夏莹[1,2] 朱小燕[1,2] 

机构地区：[1]清华大学计算机科学与技术系 [2]智能技术与系统国家重点实验室

出　　处：《清华大学学报（自然科学版）》1997年第3期42-45,共4页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金;国家"八六三"高科技项目

摘　　要：方向线素特征在手写体汉字识别中得到了广泛的应用，文中对该特征进行了形式化描述，分析了其存在的不足，在普通方向线素特征的基础上，引入模糊数学思想，提出了汉字的模糊方向线素特征，给出了形式化描述，应用该特征实现了一个非特定人手写体汉字识别系统，经对含有３７５５０个汉字（１０套国标一级汉字）的未学习样张的测试，识别率由８５．５％提高到８９．７％，十选正确率达到９８．４％，在同等测试条件下。The directional line element feature has been applied in the recognition of handwritten Chinese characters widely. This paper presents a formalized description of this feature. Basing on the common directional line element and according to its defect, the fuzzy mathematical idea is introduced and fuzzy directional line element feature is proposed . Then, a system using this feature is implemented. After testing on samples not having been learned,which contain 37550 characters (10 sets of Chinese characters specified in the GB2312 80 level I standard), recognition rate can be improved from 85.5% to 89.7%, and the correct recognition rate of 10 candidates can be reached to 98.4%. By this method, we get 4 percentage points of improvement upon using only the directional line element feature under the same conditiions.

关 键 词：汉字识别 特征抽取 手写体汉字 方向线素特征 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]