摘要
文章研究了手写体汉字识别中的一种新的特征提取方法——小波分形分解特征。在该方法中,对手写汉字分别采用小波和分形的方法提取其结构特征和统计特征,并将提取的结构特征和统计特征组合后作为识别器的输入进行识别。实验表明,该方法对训练样本可以达到98.71%的识别率,对测试样本可以达到91.37%的识别率。
This paper study a new feature extraction approach for computer recogn ition of handwritten Chinese character. In this method, wavelet and fractal were used to pick up the structure feature and statistical feature, these features w ere combined and used for recognizer as input data. The experiment show satisfie d result. Identify rate could reach 98.71% for training stylebook and 91.37% for testing stylebook.
出处
《微电子学与计算机》
CSCD
北大核心
2004年第8期115-117,120,共4页
Microelectronics & Computer
关键词
汉字识别
特征提取
小波
分形
Chinese characters recognized,Feature extraction,Wavelet,Fractal
