摘要
粗分类是提高汉字识别速度的主要手段。将RBF(Radial Basis Function neural network)神经网络用于汉字粗分类,采用汉字四边码和粗网格作为汉字粗分类的特征以进行比较。分别对GB2312-80一级字库印刷体及手写体进行实验,实验结果表明将RBF神经网络用于汉字粗分类比通常使用的欧式距离作为分类器有较好的性能。
Coarse classification is the key to improve recognition speed.An improved coarse classification scheme based on RBF （Radial Basis Function） neural network for Chinese character is presented in this paper.Four-side code feature and gross meshed feature are respectly applied as coarse feature to compare in this experiment.The GB2312-80 first-level Chinese character sam- pies including printed and handwritten form are objects in this experiment.Experiment results show that proposed method has ex- cellent performance on coarse classification in contrast to Euclidean distance as classifier used in conventional method.
出处
《计算机工程与应用》
CSCD
北大核心
2009年第6期170-172,共3页
Computer Engineering and Applications
基金
国家自然科学基金
河北省教育厅科学技术研究重点项目
河北省科学技术研究与发展计划项目
河北大学自然科学基金资助项目~~
关键词
RBF神经网络
粗分类
四边码
粗网格
Radial Basis Function（RBF） neural network
coarse classification
four-side code
gross meshed feature
