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.
出处
《微计算机信息》
2009年第12期294-295,共2页
Control & Automation
关键词
组合特征
BP人工神经网络
脱机手写数字识别
Combination feature
BP artificial neural network
Off-line handwritten digit recognition
