摘要
本文研究手写体汉字识别中的一种新的特征提取方法—弹性网格方向分解特征，在该方法中，我们将手写体汉字进行横、竖、撇、捺四方向的分解，然后根据该汉字的笔划分布构造一组非均匀的弹性网格，将弹性网格分别作用于汉字的四方向分量上，统计汉字象素点在网格中的概率分布来作为特征。对１０３４类汉字共５１７００个手写体汉字样本的实验我们得到了９４３９％的识别率，表明该特征提取方法是十分可行和有效的。
:In this paper,we propose a new feature extraction approach for computer recognition of handwritten Chinese character.The approach consists of three steps:firstly,elastic meshing technique is employed for a given handwritten character to obtain a set of elastic meshes.Secondly,the character is decomposed into several directional patterns.Finally,the elastic meshes are applied to each of the directional pattern and the probability distribution of pixels within each mesh is computed as the features for this character.Experiment on total 51700 samples of 1034 categories of Chinese characters produced a result of recognition rate of 94.3%.
出处
《电路与系统学报》
CSCD
1997年第3期7-12,共6页
Journal of Circuits and Systems
基金
国家自然科学基金
华南理工大学自然科学基金
关键词
模式识别
中文信息处理
特征提取
弹性网格
:Pattern recognition,Chinese language processing,Feature extraction,Elastic meshing technique.
