Uniform Segmentation in Online Signature Verification

In this paper, an online signature verification system based on the segmentation of online signature is proposed. The two set of features i.e shape features and dynamic features are extracted and a feature vector concatenating these two set of features is obtained. Euclidean distance is used as a classifier and a writer dependent threshold is used for verification purpose. In order to evaluate the effectiveness of the proposed online signature verification system, several experiments are carried out. Public database SVC 2004 is used for all the tests for online signature verification. The accuracy of proposed system with ten signatures in the training set and ten segments per signature is 91.75 % for skilled forgery. Experimentation is carried out separately for skilled forgery and random forgery. Experimental results on SVC2004 Task 2 show good performance of the proposed online signature verification system.