The Improvement And Application Of Hopfield Neural Network

Off-line handwritten digit recognition has wide applicative foreground in many domains, and scholars inside and outside have done much research work on it. They have reported many preprocessing algorithms and pattern recognition algorithms, which improves the accuracy of handwritten digit recognition in great measure. But up to now, the recognition accuracy still need to be improved and the problem of selecting kernel functions and kernel parameters still need to be solved.To improve the accuracy of handwritten digit recognition, this paper applies Hopfield neural network to handwritten digit recognition and exploits a software system based on Hopfield neural network."Energy function"of Hopfield neural network has the characteristic of reducing gradually and tending to be steady state of balance in the end. Furthermore, once the network is established, it can automatically move. The recognition is divided into two stages: the training stage and the recognizing stage. At the training stage, the network model is established by taking the target vector the same as the input one. At the handwritten digit recognition stage, the need-to-be recognized number features are put into the network to operate. When the network stands in balance, the output vector is compared with the vector in database, and the handwritten number with the smallest distance is the sample image.The process of handwritten digit recognition involves preprocess, feature extraction and segmentation and recognition of handwritten digit string. Firstly handwritten digit image is preprocessed through the methods, which are reducing the noise, making the image smooth and binary methods. The paper proposes a new reducing-noise technology based on the wavelet transform which will improve the accuracy and efficiency of the median filtering algorithms and will more adapt to the follow-up Hopfield neural network training and identification process. Then, the paper put forward the recognition-based Method to solve the segmentation problem; design a classifier based on the Hopfield neural networks to recognize the handwritten number, to get classifier with better performance, negative data must be the necessary trained samples and the rate of negative data and positive data must be reasonable. The classifier gets higher recognition accuracy through the increase of refuse rate.The experiment results show that the classifier trained based on Hopfield neural network with negative data and positive data get lower rate of misrecognition and better recognition accuracy.