摘要
本文在分析手写汉字识别的几种非线性归一化方法基础上,提出了五种新的基于笔画密度的弹性网格构造方法,并将之应用到手写汉字的弹性特征提取.该方法既兼顾了笔画密度对不同书写风格笔画不规则变形的适应能力,又避免了进行非线性归一化产生的笔画粗细不均匀,且计算量相对减少.针对1034类别的手写汉字样本的对比实验表明,本文方法的识别率较非线性归一化方法平均增加4.02个百分点,显示了弹性网格较强的适应笔画书写变形的能力.
In this paper, a new elastic meshing feature extraction approach based on the stroke density is proposed for handwritten Chinese character recognition. Based on the different stroke density definitions, five kinds of elastic meshes are presented. The method can not only absorb the variations of strokes in different handwritings, but also avoid the unnatural and irregular width of character strokes that often occur in nonlinear shape normalization. Experimental results on 1034 categories of handwritten Chinese characters show the effectiveness of the proposed method and indicate that it can improve recognition rate by an average of 4. 02% among different stroke density definitions compared to the nonlinear shape normalization.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2002年第3期351-354,共4页
Pattern Recognition and Artificial Intelligence
基金
国家自然科学基金(69802007)
广东省自然科学基金(980602)
Motorola研究基金
关键词
笔画密度
特征提取
手定汉字识别
非线性归一化
弹性网格
Handwritten Chinese Character Recognition, Nonlinear Shape Normalization, Elastic Meshes