基于BP神经网的手写体英文字符识别研究  

THE RESEARCH ON HAND-WRITING ENGLISH CHARACTER RECOGNITION BASED ON BP NEURAL NETWORK

作　　者：黄晓明[1] 严洪范 

机构地区：[1]上海市计算技术研究所

出　　处：《计算机应用与软件》1992年第3期12-15,共4页Computer Applications and Software

摘　　要：本文应用反向传播神经网(BP网)对大写英文字母进行识别。我们共选择了260个样本,其中一半用于训练,另一半用于识别。实验结果表明,BP网比传统的分类器(NN分类器)具有更高的识别率,其正确率达97%。另外,本文对BP网的参数与识别准确率、学习速度的关系进行了讨论,并给出了它们的关系曲线。The capital English characters are recognized by the back propagationneural network(BP network)in this paper.Of selected 260 samples,half is for trai-ning and anothers one is for recognizing.According to the experiment result,theaccuracy of recognition of BP network is greater than that of coventional classifiers(NN classifiers).The accuracy is 97%.We also discuss the relations among parame-ters of BP network,accuracy of recognition and learning speed,and give their relatedourves.

关 键 词：字符识别 手写体 神经网络 英文 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]