The Research On Handwritten Chinese Characters Recognition Based On Deep Learning

Chinese characters is used by most people in the world,the most concise information storage,the most widely used language,is an indispensable part of China traditional culture and world culture,so for the identification of Chinese characters image has very important practical value.Handwritten Chinese character recognition in daily office automation,mail sorting,finance and other fields have very good prospects,handwritten Chinese characters recognition research can make people's daily life more convenient.But because of the Chinese characters category,and the structure is complex,there are a large number of similar Chinese characters,and different people have different style of writing,writing,writing environment in different situations,the image of the handwritten Chinese characters from the types and similarity are very complicated,which makes the handwriting recognition Chinese characters in pattern recognition the field has been difficult and hot research.In recent years,depth learning has become the most popular research field in the field of machine learning,especially in the field of image recognition.It can express complex function in a more concise way,can automatically obtain sample probability distribution and has advantages in learning sample characteristics.Therefore,the depth learning model is applied to the off-line handwritten Chinese characters recognition task.The accuracy of handwritten Chinese characters recognition is further improved by using different methods and models in deep learning.This paper mainly includes the following aspects:Firstly,starting from the deep belief network,this paper analyzes the shortcomings of traditional handwritten Chinese characters recognition method,and proposes a method based on deep belief network fusion model for Chinese character recognition.This method first uses modified quadratic discriminantfunction classifier simple Chinese characters correction,and Chinese characters complex image to a deep belief network model,the specific process of division of labor through the definition of credibility to coordinate the division of two classifiers in a recognition task,so as to achieve good results.Then from the convolution neural network,the convolution neural network is superior to the traditional image classification method in feature extraction,so the convolution neural network is applied in the off-line handwritten Chinese character recognition problem.Analysis of the convolution in the processing of handwritten Chinese characters and disadvantages,for similar handwritten Chinese characters classification problems,to further improve the convolutional neural network,making full use of network advantage feature extraction,classifier two using SVM classifier classic,the combination of better recognition and classification of similar handwritten Chinese characters.In this paper,MNIST handwritten data collected by the Institute of automation,Chinese Academy of Sciences,CASIA-HWDB1.1,Beijing University of Posts and Telecommunications collected HCL2000 data set.Experiments show that the proposed handwritten Chinese character recognition method based on convolution neural network can achieve better recognition effect.