Handwriting Verification Using Point and Side Information for Chinese Words

In this manuscript, we proposed a verification algorithm for the Chinese word. We use a variety of different features including global features and local features. Considering that Chinese words often have a more complicated structure, we adopt global features, matched side feature, and matched point feature, which are features extracted from different magnitude levels. The three feature sets are able to describe a Chinese word in a more complete way. Support vector machine is used as a classifier. Experimental results show that the proposed algorithm reaches 95.85% accuracy in handwriting verification and outperforms other methods.