摘要
该文提出一种基于模糊模型相似测量的文本分析系统的字符预分类方法 ,用于对字符的无监督分类 ,以提高整个字符识别系统的速度、正确性和鲁棒性 .作者在字符印刷结构归类的基础上 ,采用模板匹配方法将各类字符分别转换成基于一非线性加权相似函数的模糊样板集合 .模糊字符的无监督分类是字符匹配的一种自然范例并发展了加权模糊相似测量的研究 .该文讨论了该模糊模型的特性、模糊样板匹配的规则 ,并用于加快字符分类处理 ,经过字符分类 。
This paper presents a character preclassification method based on similarity measure in Fuzzy model to perform unsupervised character classification for improvement in robustness, correctness, and speed of a character recognition system. On the basis of character typographical structure categorization, a pattern matching is used to classify the characters in each category into a set of fuzzy prototypes based on a nonlinear weighted similarity function. The emphasis of inequality measure for small characters guarantees no misclassification, but a little redundancy is encountered on the fuzzy prototype set. This redundancy can be removed by self grouping of the final prototype set. The fuzzy unsupervised character classification, which is natural in the representation of prototypes for character matching, is developed and a weighted fuzzy similarity measure is explored. A fuzzy model of prototypes is defined and several propositions of the features of the fuzzy model are given. The characteristics of the fuzzy model and rule based matching of fuzzy prototypes are discussed and used in speeding up the classification process. The fuzzy model of prototype has been verified to reduce the effect of noise. Based on prototypes that are free of noise, the recognition problem will be simplified and the speed as well as recognition rate will be increased. For ambiguous characters, probably as merged, the accuracy of postprocessing also will be improved. After classification, the character recognition, which is simply applied on a smaller set of the fuzzy prototypes, becomes much easier and less time consuming.
出处
《计算机学报》
EI
CSCD
北大核心
2002年第4期423-429,共7页
Chinese Journal of Computers
基金
国家自然科学基金(7870 0 12 )
江苏省教委留学回国人员科研基金(199715 5 1)
江苏省教委自然科学研究基金 (99KGB14 0 0 0 9)资助
关键词
模糊模型
加权模糊相似测量
字符无监督分类
匹配算法
分级归类
字符识别
字符匹配
fuzzy model, weighted fuzzy similarity measure, unsupervised character classification, matching algorithm, classification hierarchy
