摘要： 通过对基于概率密度函数估计的直接方法和基于样本间相似度量的间接方法的结合使用,并针对手写体汉字识别的特点,对相应算法及过程进行改进,使经非监督学习后的样本集的划分达到整体和局部的最佳。

关键词: 概率密度, 训练数据, 子集, 样本集, 样本模

Abstract: Aiming at the characteristics in Chinese characters manuscript recognition, this paper improves corresponding algorithm and process by combining probability density function estimating based direct method and makes the division of sample set after non-monitored learning get to integral and partial optimization. ;;;;

中图分类号: 

• TP391.4