Research On Recognition Method Of Handwritten Numbers And English Characters

The pattern of making online emerged in recent years has improved many disadvantages of traditional work such as Marking intensity, and this advanced Marking model has been applied to various types of standardized tests in China. The most representative system is based on OMG technologies. But this kind of system cost too much to promoting. Therefore, another kind of system based on image processing technology emerged.This paper is to study the recognition method of handwritten numbers and character in online making system which based on image processing technology. Researchers usually put together with the number and character study, it is because there is no essential difference between the two object. From the perspective of identification methods, there is only difference of test samples, there is no difference in research on the identification methods. Therefore, in order to facilitate, we only discuss in name of methods of number identification .In order to improve the accuracy rate of recognition of handwritten number, the researchers have done a lot of work at the pre-processing stage, feature extraction and classification stage. Looking at all kinds of methods, they have their own advantages also have their own deficiencies which they can not overcome. Currently, the hottest method is neural network method and support vector machine method which also have the highest accuracy rate. BP neural network has been done lots of work among Neural Networks methods, which also has a higher recognition accuracy in the same kind of methods, but it cost the long training time, and is easy to fall into local minimum. And Support Vector Machine method is more adept at classification problems of two types.In this paper, the author use RBF neural networks and support vector method to build two-level classification model. The first classifier using RBF neural network structure is allowed to output the final result, fuzzy result, and the samples which the first classifier can not handled. If the output of the first classifier is the fuzzy result or the sample it can not handled, system will switch to the second classifier, all that is entered into SVM classifier to continue to identify and output the final result. So a joint multi-classifier system, certainly can greatly improve the system recognition rate.