Offline Handwritten Chinese Character Recognition Based On Deep Learning

Chinese characters are the most widely used words,carrying the long history of culture in our country,has an influence on our lives.Offline handwritten Chinese characters in the field of modern office,financial,intelligent and has great application prospect.But offline handwritten Chinese characters as category huge number,number of strokes and similar words,the process of writing and to personal writing habit,writing environment,writing style,the influence of such factors as the difficulty and hotspot in the field of pattern recognition.In recent years,with the rise of the deep learning,its excellent feature extraction performance and recognition performance have attracted wide attention of scholars at home and abroad.It has been applied in many fields and achieved good results.The recognition rate of offline handwritten Chinese characters and their similar characters is further improved by using deep learning in this thesis.The specific contents are as follows:1.Aiming at the problem of low recognition rate of offline handwritten Chinese characters,this thesis proposes a fusion model based on modified quadratic discriminant function and deep Boltzmann machine.The main idea of this model is that modified quadratic discriminant function and deep Boltzmann machine in feature extraction and classification mechanism can complement each other.The model recognizes simple Chinese characters by modified quadratic discriminant function.Only a small number of Chinese characters whose large probability will be misclassified are handed over to the deep Boltzmann Machine for recognition,and the division of labor is coordinated by defining the generalized confidence degree,which effectively overcomes the weakness of the modified quadratic discriminant function in recognizing complex characters and the high computational complexity of the deep Boltzrmann machine,and significantly improves the recognition rate of offline handwritten Chinese characters.2.Analyzing the shortcomings of current research on similar offline hand-written Chinese characters,a cascade model based on deep belief network and support vector machine is proposed to solve the problem of difficult recognition.The model first extracts the features of offline handwritten Chinese characters using deep belief network with excellent feature extraction performance,and then generates similar offline handwritten data sets based on frequency statistics.Finally,strong classifier support vector machine is used for recognition and classification.The experimental results show that the cascade model proposed in this thesis effectively improves the recognition effect of similar offline handwritten Chinese characters.