Abstract
Just like its remarkable achievements in many computer vision tasks, theconvolutional neural networks (CNN) provide an end-to-end solution inhandwritten Chinese character recognition (HCCR) with great success. However,the process of learning discriminative features for image recognition isdifficult in cases where little data is available. In this paper, we propose anovel method for learning siamese neural network which employ a specialstructure to predict the similarity between handwritten Chinese characters andtemplate images. The optimization of siamese neural network can be treated as asimple binary classification problem. When the training process has beenfinished, the powerful discriminative features help us to generalize thepredictive power not just to new data, but to entirely new classes that neverappear in the training set. Experiments performed on the ICDAR-2013 offlineHCCR datasets have shown that the proposed method has a very promisinggeneralization ability to the new classes that never appear in the trainingset.