基于组合分类器的联机手写汉字识别  

On-line Chinese Character Recognition Based on 2-classifier Combination

作　　者：郑桢[1] 康耀红[1] 

机构地区：[1]海南大学信息科学技术学院,海口570228

出　　处：《郑州大学学报（理学版）》2007年第2期88-91,99,共5页Journal of Zhengzhou University:Natural Science Edition

基　　金：教育部重点基金资助项目;编号03144;海南省自然科学基金资助项目;编号60533

摘　　要：提出一种基于二级组合分类器的汉字手写体识别方法,首先提取汉字的整字特征和笔划特征;进行单字识别的过程中,GA-BP笔划分类器起着对汉字识别的主导作用,当写入汉字出现连笔、字形变化导致系统产生拒识情况时,采用隐马尔可夫整字分类器作为辅助去提高整体的识别率.The method proposed is based on 2-classifier combination. Firstly the whole features and the strokes features are extracted from the input Chinese characters. During the recognition of single character, GA-BP neural network based on stroke classifier is used as a primary tool to recognize the Chinese characters. But if the words written by consecutive and metamorphic strokes cause the exclusion to appear unexpectedly, HMM based on whole classifier is introduced to improve the whole recognition rate.

关 键 词：二级组合分类器 整字 笔划 GA-BP网络 隐马尔可夫模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]