基于一种新的组合特征的脱机手写数字识别  

Off-line Handwritten Digit Recognition Based on a new combination feature

作　　者：张承德[1] 黄襄念[1] 张凯兵[2] 卢华[1] 

机构地区：[1]西华大学数学与计算机学院,四川成都610039 [2]孝感学院计算机科学与技术系,湖北孝感432100

出　　处：《微计算机信息》2009年第12期294-295,共2页Control & Automation

摘　　要：本文提出了一种新的组合特征。通过对方向线素特征、轮廓特征和有效行特征进行有效组合形成一组新的组合特征。该新的组合特征无需对样本进行细化等操作,减少了前期的预处理工作量,也减少了因细化变形而引起的误识,用BP人工神经网络作为分类器,实验表明效果良好。This paper has presented a new combination of features The new combination feature is composed of directional element features, Contour Features and real line feature. This new method has little operation about thinning of the samples, so it can not only decrease the works of pretreatment, but also reduce the error caused by thinning, and the BP artificial neural network was used as the classifier. The experiments means： it is a good method.

关 键 词：组合特征 BP人工神经网络 脱机手写数字识别 

分 类 号：TP[自动化与计算机技术]