Preprocessing for handwritten Chinese character based on neural networks

This paper introduce entropy concept to backpropagation neural network in order to improve its performance of classifing similar sample. Then,based on the neural network,we propose a thinning model for Chinese character, and state its preprocessing methods, neural network architecture,learning algorithm and working algorithm. At the same time,we reform the conventional thinning process for Chinese characters in order to improve speed of processing. Finally, cooperating the nonlinear shape normalization methods and the reformed thinning process, the experiment on the handwritten Chinese character preprocessing has been done to test the model,and obtain the satisfactory results.