基于双弹性网格的手写体汉字识别

特征提取是手写体汉字识别的关键，目前四方向网格特征已被实验证实是一种较好的手写体汉字特征。针对通常的纵横弹性网格对汉字“撇、捺”笔画特征提取的不足，提出一种新的网格构造技术——对角弹性网格，它由45°和135°的对角直线构成，将汉字图像划分为多个菱形，能够很好地适应汉字在“撇、捺”方向的变化。将这两种网格单独，以及相互组合成双网格等情况分别进行手写体识别实验，实验结果验证了对角弹性网格的有效性和双弹性网格的高识别率性。

Handwritten Chinese character recognition based on double elastic mesh

Abstract: Feature extraction is the key part of handwritten Chinese character recognition. It is found that the directional feature digged by elastic mesh is suitable for Chinese character recognition. In view of the usual vertical-horizontal elastic mesh’s shortage in left falling stroke and right falling stroke, we present a new elastic mesh structure method—diagonal elastic mesh, which is composed of diagonal beelines of and , and divides the Chinese character image into many diamonds. This paper takes the two kinds of mesh and their combinations into the handwritten recognition experiment. The experimental results have confirmed the diagonal elastic mesh’s validity and that the double elastic mesh has very high recognition rate.