摘要： 信息的连续采集会造成部分字符存在连笔，进而影响字符识别率。为此，提出一种基于连笔消除的空间手写字符识别方法。将空间手写字符平面化，提取字符拐点和笔画方向特征。为避免笔画的误消除，利用支持向量机把未知字符分为带连笔字符和非连笔字符，通过连笔的书写特征消除连笔，将空间字符轨迹转化为平面字符轨迹，直接用平面字符分类器进行字符识别。实验结果表明，该方法连笔消除效果显著，利用现有字符库即可获得较高的字符识别率。

关键词: 支持向量机, 三维空间手写, 特征提取, 笔画分段, 连笔消除

Abstract: During the process of continuous collection of information in 3D space handwriting, connected strokes of part of characters appeares, which affectes the rate of characters recognition. To solve this problem, an approach of 3D space handwriting character recognition based on elimination of connected stroke is represented. 3D space characters are flattened. Feature extraction of turning points and direction of strokes is performed. To avoid eliminating the wrong strokes, Support Vector Machine(SVM) is adopted to divide the unknown characters into 2 sections: connected-stroke characters and non-connected-stroke characters. The connected strokes are eliminated by using stroke writing features which put the trajectory of a 3D space character into a flat trajectory of the character. Character recognition is performed by using flat character classifier directly. Experimental results show that the effect of eliminating connected stroke using this approach is remarkable and a higher recognition rate can be obtained by using existed character library.

Key words: Support Vector Machine(SVM), 3D space handwriting, feature extraction, stroke segmentation, elimination of connected stroke

中图分类号: 

• TP212.9