Research On Handwritten Chinese Character Recognition Technology Based On Deep Learning

With the innovation of knowledge culture and the development of science and technology,people's requirements for daily life quality standards have been improved,and the requirements for electronic products have become more and more demanding.New requirements for handwritten Chinese characters have been put forward,and handwritten Chinese character recognition is processing documents.There are huge potentials in social activities such as the classification of mail,business,and so on.However,the large amount of Chinese characters,the complicated structure,the near-words and the changing writing styles make the Chinese character recognition technology more difficult,and the existing solutions are difficult to reach the level required by people.It seems that as machine learning is gradually becoming known to researchers,the path of deep learning has also expanded and become the most popular research content in the mechanical sciences in a short period of practice.Deep learning can automatically obtain the sample probability distribution by simply displaying the complex function,which shows an extraordinary advantage in the learning sample features.Researchers mainly apply it to the field of image recognition.Therefore,based on the convolutional neural network,LeNet-5 structural model and TransFlow framework,the CASIA-HWDB1.1 handwritten Chinese character dataset collected by the Institute of Automation of the Chinese Academy of Sciences was implemented by Python programming software.The first two chapters of this study mainly describe the background,purpose and significance of the research.Then,through the literature research method,the research status of handwritten Chinese character recognition technology based on deep learning at home and abroad is sorted out,and the difficult problems of the research are analyzed.The Chinese character recognition technology is divided into two parts.The former part is online handwritten Chinese character recognition.The computer recognizes the stroke order,stroke direction and Chinese character shape of the input Chinese characters.The current development is mature and in the computer.It has a wide range of applications;the latter part is offline handwritten Chinese character recognition.Due to the large number of Chinese characters,complicated layout,multiple similar words and varied styles,the recognition rate of Chinese characters isstill low,and the main content of this study is briefly explained.This research mainly uses convolutional neural network as the research foundation,and introduces the related concepts,development history and common methods of deep learning.The third chapter is the core content of this research.Firstly,the design of handwritten Chinese character recognition technology based on deep learning is designed,including the process design and system design of handwritten Chinese character recognition based on deep learning.Then is the specific process of the experiment,according to the design process,first set up the experimental platform and then determine the handwritten Chinese character training data set and test data set to complete the experimental preparation work.Next,normalize the sample data image,smooth denoising and affine transformation,and put the processed sample image into the LeNet-5 structural model based on convolutional neural network for learning and training.Multiple advanced APIs in the TransFlow framework are programmed.Finally,specific experiments are carried out in the established platform environment to realize the function of the LeNet-5 structural model.The recognition results of handwritten Chinese characters are obtained,and the experimental results are further analyzed.The study summarizes and looks at the entire study.