Study On Selection For SVM Kernel Parameter In Handwritten Chinese Character Recognition

With the rapid development of IT, handwritten Chinese character recognition has been widely used in file retrieval, office automation, mail system and bank bill process, etc. Because it has too many technology difficulties, large Chinese character set recognition still can not achieve real-time processing requirement. But in a variety of applications, only small character set is needed. For example, there is an urgent demand for the recognition of handwritten financial Chinese character in financial industry. So the research on small character set is of great value, and also can guide the research on recognizing large Chinese character set.Financial Chinese character belongs to Chinese character. So this paper introduced the procedure of Chinese character recognition, and discussed the design methods of classifier in detail. Because SVM has the particular advantages in fine classification for small character set, this paper used SVM classifier to recognize financial Chinese character.SVM is a kernel-based method, kernel function and kernel parameter selection directly affect SVM model's generalization ability. After analyzing the features of SVM kernel parameter and the methods of estimating generalization ability deeply, this paper proposed using the separability measure between classes in the feature space to choose the kernel parameter. Calculating such SM costs much less computation time than training the corresponding SVM models, so the best kernel parameter can be chosen much faster, and the testing accuracy of trained SVM by the proposed method is competitive to the standard ones. Experiment results show that the proposed method can effectively improve classification speed of handwritten financial Chinese character and get better recognition accuracy.