基于多信息融合的自然书写脱机中文文本行识别  

Recognition of Unconstrained Handwritten Offline Chinese Text Line Based on Multiple Information Fusion

作　　者：李南希[1] 金连文[1] 

机构地区：[1]华南理工大学电子与信息学院,广东广州510640

出　　处：《华南理工大学学报（自然科学版）》2010年第12期45-49,共5页Journal of South China University of Technology(Natural Science Edition)

基　　金：国家自然科学基金重点项目(U0735004);国家自然科学基金资助项目(60772116);广东省自然科学基金重点项目(07118074)

摘　　要：自然书写的脱机中文文本行识别是当今字符识别领域的一个难题.为降低文本行识别中负样本的干扰,文中提出了一个概率模型,将负样本作为一种信息来源,与单字符的识别信息、文本行的几何信息等进行融合.简单地使用两个分类器,就可以实现上述概率模型.在多人手写真实文本数据库中进行的实验表明,当无语言模型和使用二元语言模型时,文中所提方法的正确识别率分别达到61.29%和72.73%,体现了该方法的有效性.The recognition of unconstrained handwritten offline Chinese text line is a difficult problem in current character recognition domain.In order to reduce the interferences from negative samples during text line recognition,a probabilistic model is developed.In this model,negative samples are treated as an information source and are integrated with other kinds of information such as the recognition information of isolated characters and the geometric information of text line.By using only two classifiers,the proposed probabilistic model can be successfully implemented.Experimental results on a real multiple-writer handwritten text database show that,by using the proposed method,the correct recognition rates reach 61.29% without any language model and 72.73% with a bi-gram language model,respectively,which means that the method is effective.

关 键 词：字符识别 中文文本行识别 多信息融合 概率模型 负样本 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]