手写字符识别中的特征提取研究  

Research on the Feature Extraction of Recognition Handwritten Characters

作　　者：胡龙灿[1] 孙怀义 樊爱军 

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065 [2]重庆市科学技术研究院信息与自动化技术研究中心,重庆401123

出　　处：《自动化与仪器仪表》2013年第2期15-17,共3页Automation & Instrumentation

基　　金：重庆市科技攻关项目(CSTC201140003;CSTC2011GGB40033;CSTC2012gg-yyJSB40004);国家科技支撑计划课题(2012BAH55F01)

摘　　要：针对联机手写字符的特点,有效地将脱机手写字符的结构特征和统计特征与联机手写字符的笔划特征联系起来,使用内部圆个数,宽高比,密度,射线个数,线段,尖角,斜率七种特征形成模板,使用贝叶斯网络进行培训并识别。实践证明了该方法的可行性。Due to the characteristic of on-line, the structure and statistical characteristic of handwritten characters off-line are combined effectively with the stroke characteristics of handwritten characters on-line, the template is formed with seven characteristics, including the number of internal circles, aspect ratio, density, the number of ray, segment and slope, and Bayesian network is used to train and recognize. Practice has proven that the method is feasible.

关 键 词：字符识别 特征提取 贝叶斯网络 

分 类 号：TP273.4[自动化与计算机技术—检测技术与自动化装置]