Automatic Chinese Handwriting Verification Algorithm Using Deep Neural Networks
Chi-Chang Lee, Jian–Jiun Ding · 2019
Handwriting verification is to identify whether a script was written by a person himself or forged. Conventional handwriting verification algorithms are based on feature extraction. However, the features of scripts are highly affected by the writing instrument, the posture, and the force of writing, even if the scripts were written by the same person, the extracted features will be quite different. Moreover, since some writers might not write some strokes clearly or ignore some strokes, not all features can be well extracted in every script. Therefore, in this paper, we apply a deep neural network based algorithm for handwriting verification. With the proposed algorithm, the parts that are really powerful and robust for handwriting verification can be highlighted by the auto-encoder. Then, a very high accurate handwriting verification result can be achieved.