The Research Of Off-line Handwritten Chinese Character Recognition Based On BP Neutral Network

Off-line Handwritten Chinese Character Recognition is one of the most favorable areas in the pattern recognition, and is a difficult project in computer character recognition too. After analyzing the currently up-to-date techniques of Chinese character recognition, we discussed the Chinese character recognition of the small category character. In order to improve the accuracy and speed of the recognition, we gave some improvements of the BP neutral network used for the small category character recognition. The mainly work of this paper includes:First, we studied the preprocessing of the handwritten recognition, including standardization, separation, erasing the redundancy information and thinning it.According to the feature of the small category character, this paper used the method of elastic grids and direction stroke of pixels to extract the feature vectors, which are taken as the inputs of the BP neutral network. In the designing of recognition classifier, we discussed the drawbacks of the BP neutral network in small category character recognition and made improvements of it. This paper mainly discussed the structure of BP neutral network, the number choice of nerve cells and the parameters of the improved BP neutral network, etc.We discussed the realization of BP neutral network on MATLAB, and presented the foundation, training and simulating of the improved BP neutral network.In this paper, the four Chinese characters'优,良,中,差', which are often used in Teacher's handwritten Chinese character grades recognition, are used for the simulation experiments. The experimental results show that the improved BP neutral network methods, in the small category character recognition, can improve the accuracy and speed of the recognition.