The Generation Of Handwritten Chinese Characters Based On The Generation Confrontation Network

Chinese characters accompany the long history of Chinese culture and are a symbol of the glory of Chinese culture.Over the past three thousand years,more than one-quarter of the world's people have used Chinese characters.Chinese characters have been one of the basic tools for education,employment,communication,and daily communication in China and even East Asia.In addition,Chinese characters are not only used as a language.Since ancient times,many calligraphers have left their masterpieces.Chinese calligraphy is not only a language expression,but also a visual art.With the advent of the digital information age,the application scenarios and demands of handwritten Chinese characters have gradually increased,and it has become particularly important to use standard fonts to generate handwritten fonts.Traditional handwriting font generation methods mostly focus on the decomposition of strokes of Chinese characters and the establishment of hierarchical relationships,and do not focus on the overall style of Chinese characters.Generative Adversarial Networks(GAN)have created new ideas for this.GAN can learn the sample distribution of known real data,and then learn the sample and generate true and false data.For the generation of handwritten Chinese characters,not only to generate the Chinese characters prepared by strokes,but also to learn the writing style of Chinese characters,the general GAN can not complete such tasks.In order to better complete the task of handwritten Chinese character generation,this paper proposes to generate handwritten Chinese characters based on improved Cycle GAN.Cycle GAN is one of the derivative models of GAN in recent years.In this paper,by improving the network structure of the transmission module in the generator,the residual network(Res Net)in the original structure is replaced with a dense cascade network(Dense Net),which encourages the generation of the confrontation network to learn more detailed information,so as to better Generate images of handwritten Chinese characters.Image pre-processing and feature extraction are important means of image recognition.In this paper,the generation of Chinese character images generated by the anti-network is used as the data set of the handwritten Chinese character recognition experiment.From the accuracy of Chinese character recognition,the effectiveness and feasibility of this method is measured.The recognition rate of the Chinese character image generated based on the method in this article is higher than that of the original method under different features.At the same time,based on the most intuitive human vision,the Chinese character image generated in this paper has a more complete Chinese character structure,clearer outline,and more beautiful.