Deep Learning and Lexical Analysis Combined Rubbing Character Recognition

Rubbings are among the oldest ancient literatures and presumably contain a lot of knowledge yet to be discovered. However, as of now, the recognition process of rubbing characters is now completely done manually. Such process, due to varying styles of characters, takes an enormous amount of effort, and can be done by a small number of specialists. Still, the aging process, which may damage characters or add noises, further increases the difficulty of recognition. In this paper, we propose a method by combining deep learning and lexical analysis for increasing the accuracy of rubbing character recognition. This work is currently in progress, and the experimentation uses rubbing character database developed by Kyoto University. In the experiment, we use a rubbing image for recognition, and the experimental results verify the effectiveness of the proposed method.