摘要
针对常用的银行汉字和阿拉伯数字混合字符集的识别,提出了依据不同的分类要求,分别选取不同的分类特征,并采用先聚类再用多层感知器(MLP)神经网络分类的多级分类器进行识别的设计方法。实验结果表明,该方法用于手写体混合字符集的识别是行之有效的。
A novel method for the recognition of commonly used handwritten Chinese characters in bank and digits was presented. Considering diverse classification requirements, different features were selected, an integration of clustering and Multi-layer Pereeptron neural networks was utilized. The experiment demonstrates that the proposed approach is promising.
出处
《计算机应用》
CSCD
北大核心
2005年第12期2948-2950,共3页
journal of Computer Applications
关键词
手写体字符识别
特征选取
多级分类器
handwritten character recognition
feature selection
multilayer classifiers