一种手写汉字特征抽取的新方法

A New Feature Extraction Method for Off-line Handwritten Chinese Character Recognition

针对非特定人脱机手写汉字识别，本文提出了一种新的特征抽取方法。首先，将输入汉字模式进行非线性规一化；然后，根据汉字的横竖撇捺四种基本笔画将规一化后的模式分割成四个子模式；最后，将所得到的四个子模式分别均匀地划分成M×M个小网格，在每一小网格内统计笔画穿透数目，从而得到一个4M2维的特征向量。通过对国标一级3755个汉字的测试表明，正确识别率达90％以上，从而证实了本文方法的有效性。

A new method to extract crossing line feature for off-line handwritten Chinese charactor recognition is proposed in this paper. Firstly, the input pattern is nonlinearly normalized.Secondly, the normalized pattern is separated into four subpatterns according to the four kinds of elementary strokes. Thirdly, the four subpatterns are uniformly divided into M×M cells respectively. in each cell, the crossing lines are counted. A 4M2 dimensional feature vector is generated at last. According to our experiments which test the 3755 categories of Chinese charac ters used in our daily life, the correct recognition rate is more than 90%.