Research on Offline Handwritten Chinese Character Recognition Based on Deep Learning

Offline Handwritten Chinese Character Recognition (HCCR) is not only a research difficulty, but also a research hotspot. With the rapid development of deep learning, offline HCCR which based on deep learning has been widely studied and made a Breakthrough Development in methods and performance. CASIA-HWDB1.1 handwritten Chinese character database which was collected by Institute of Automation, Chinese Academy of Sciences is adopted in the paper. In the paper, tensorflow software platform is used, and the structure of convolution neural network (CNN) is optimized. The neural network model used in the paper has been trained and tested. It can be concluded that the recognition rate of offline HCCR has been improved. The training model has been successfully applied to the off-line HCCR system based on Android, and good results have been achieved.