An Effective Writer Verification Algorithm Using Negative Samples

• Authors: Xianliang Wang;Xiaoqing Ding
• Venue: IWFHR '04 Proceedings of the Ninth International Workshop on Frontiers in Handwriting Recognition
• Year: 2004

Writer verification is the process of deciding on the claimed identity of a writer by comparing some attributes of test handwriting with those of his reference handwriting, which has great applications in forensic justification. However, automatic writer verification is very difficult, and at present most writer verification tasks are fulfilled by document examiners, which is very time-consuming and tend to be subjective. In this paper, we proposed an effective writer verification algorithm using negative samples. This algorithm uses handwriting Chinese characters as the object. In the algorithm, the directional element features are first extracted from the handwriting Chinese character scripts, then negative handwriting samples are introduced, and the decision is made based on the combination value of similarity measure between test handwriting and reference handwriting and that between test handwriting and negative handwriting. Experiments demonstrated the effectiveness of the proposed algorithm.