基于集成RBF神经网络的小类别手写体汉字识别系统  

Small Set Handwritten Chinese Character Recognition Based on Integration RBF Neural Network

作　　者：居琰[1] 汪同庆[1] 刘建胜[1] 王贵新[1] 彭健[1] 

机构地区：[1]重庆大学光电工程学院,重庆400044

出　　处：《计算机工程与应用》2002年第23期100-102,158,共4页Computer Engineering and Applications

摘　　要：该文介绍了RBF神经网络的模型,讨论了RBF网络分类器的机理和特点,提出了一种集成RBF神经网络并应用于小类别手写体汉字识别系统的设计,采用了组合重心分解网格特征方法来提取汉字特征,设计了遗传进化隐层节点自生成算法用于RBF的训练。实验表明该小类别手写体汉字识别系统有很高的识别率,具有一定的实用推广价值。This paper presents Radial Basis Function Neural Network(RBFNN),and discusses its classifying mechanism and characteristic.An integration RBFNN is used for small set handwritten Chinese character recognition.This paper pro-pose a new feature ex traction named combined barycenter mesh decomposed algorithm.Genetic evolutionary self-generat-ing of hidden units algorithm is contrived to train RBF network.Experiment of this small set handwritten Chinese char-acter recognition system possessing a high level of recognition rate,shows that the System is practical.

关 键 词：集成RBF神经网络 小类别手写体汉字识别系统 汉字特征提取 模式识别 人工神经网络 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术] TP1[自动化与计算机技术—计算机科学与技术]