Online signature verification using double-stage feature extraction modelled by dynamic feature stability experiment

What is it about?

Signature-based Identification is a kind of behavioral biometric method which is not offensive or invasive- in spite of some other methods such as fingerprint. There are some key regions in every person's signature which rarely change in various tries and are difficult to forge. In this study, Persian signers' signature behavior is found via pen-tip's velocity, pressure and acceleration profiles and applied for verification.

Featured Image

Why is it important?

- In the introduced experiment, most stable trajectories in Persian signatures are detected and have key role in the rest of the study. - To tackle a major obstacle in signature verification, radon transform and convolutional neural network in a complimentary roles established 2 level feature extraction and classification. - There are three independent path for verification. Their results are combined based on DFS experiment introduced in the paper.