摘要： 在采用支持向量机分类器识别的基础上,提出了一种基于遗传算法进行手写体相似汉字特征选择的方法,解决了相似字中部分空间的自动确定问题。采用弹性网格与小波变换相结合的方法提取相似汉字的特征。实验结果验证了基于交叉验证的特征选择方法的泛化性能好于基于简单验证的方法的结论。

关键词: 相似汉字, 特征选择, 遗传算法, 支持向量机

Abstract: A feature selection approach for similar handwritten Chinese characters recognition is presented, which is based on genetic algorithmsand support vector machines classifier. The technique of combining wavelet transform with elastic meshing is employed for a given handwrittencharacter to extract its feature. The optimal features are selected by genetic algorithms and the problem of determining partial space of similarcharacters is solved automatically. The experiment results confirm the conclusion that the generalization performance of cross-validation fitnessmeasure is better than that of in-sample validation one.

中图分类号: 

• TP391.4