摘要
本文提出了一种快速的在线手写数字识别方法,该法采用书写笔划走势时手写数字进行建模,运用决策树学习算法进行数字分类识别。数字笔划走势特征提取简单、区分度高、对用户不敏感,实现了有限的资源条件下的高速识别,同时保证了方法的良好用户适应性；决策树学习算法分类情况全面,保证了方法的高识别率。实验结果表明：该方法既具有简单高效的特点,又具备很好的用户适应性。
This paper proposed a fast method for handwritten digit recognition. The proposed method modeled the handwritten digits with the variation of stroke direction, and did the classification by decision tree learning. The extraction of stroke direction variation is simple, highly discriminable and insensitive to different users, implementing fast recognition under strict resource constraints, and the user adaptability is guaranteed at the same time. The decision tree learning is able to cover all the variance in user input, which guaranteed high recognition rate. The outcome of the experiments demonstrated the proposed method to be simple and effective with good user adaptability.
出处
《计算机科学》
CSCD
北大核心
2006年第9期207-210,F0003,共5页
Computer Science
基金
国家自然科学基金(编号:69903006
60373065)。
关键词
笔划走势
方向码
决策树
IF-THEN规则
ID3算法
Variation of stroke direction, Direction code,Decision tree, If-then rule,ID3 algorithm
