Research On Handwritten Digital Recognition System Based On Neural Network

As artificial intelligence has become a hot topic nowadays,more and more scientific researchers have devoted themselves to the field of artificial intelligence.In everyday life,characters can be seen everywhere.An important direction in character recognition is handwritten digital recognition.Borrowing from the theory of bionics,if the computer can be made to work like a human eye,the recognition of handwritten numbers can be done automatically.It will reduce the pressure of manual recognition of handwritten digital information and increase the efficiency of personnel in related industries.After a long period of development,the theory and algorithm of handwritten digital recognition have been further improved.However,how to effectively apply it in real life is still a serious problem.Neural network is a simulation of the human brain and is a manifestation of bionics.This paper studies the two network models of BP neural network and convolutional neural network.The feature extraction method has a great influence on the network recognition rate.So this paper studies three kinds of feature extraction methods,which are PCA feature,Sobel feature and Gabor feature.By studying the combination of different feature extraction methods and BP neural network,it is found that the best recognition result is obtained when Gabor features are combined with BP neural network.The feature of artificial extraction is influenced by the feature extraction method.Therefore,this paper also studies the convolutional neural network.It can realize feature extraction and identification through the network itself.For deep neural networks,the difficulty of training will increase as the number of network layers deepens.Therefore,how to use shallow network to obtain a high recognition rate is worth studying.After analyzing the LeNet5 model,this paper proposes two improved models.Through the analysis of the recognition results,It can be seen that the recognition results of different models have certain differences.Therefore,this paper proposes a combined neural network.Through experiments,it is found that the network can obtain a higher recognition rate.Finally,this paper designs a handwriting digit recognition system.The system can effectively identify handwritten digits on pictures,which has certain practical value.