Title: An efficient approach for dynamic signature recognition
Author: Sahar Abd El_Rahman
Address: Faculty of Engineering – Shoubra, Benha University, Cairo, Egypt; College of Computer and Information Sciences, Princess Nourah Bint Abdulrahman University, Riyadh, Saudi Arabia
Abstract: Online signature recognition is an accepted biometric technique, because it is less expensive than other biometric techniques. In this paper, an efficient two-stage online signature recognition approach is presented. This approach depends on the initial analyses of global features using Euclidean distance to quickly discard outlier signatures; then followed by local features analysis using an enhanced DTW algorithm. An emphasis was created to extract stroke-associated features for global recognition phase as well as for signal pre-processing prior to local recognition. Stroke-related features used in the proposed system contribute well in enhancing the run-time performance by quick discriminating genuine and forgery signatures. Experiments are run out on MCYT-100 benchmark database. The proposed system is tested with 100 users - including 25 skilled forgery signatures and 25 genuine signatures per each user. The worst FAR recorded value is less than 4%, FRR is less than 20% and EER is less than 5%.
Keywords: biometrics; dynamic signature identification; global features; dynamic time warping; DTW; local features; online signature identification; online signature recognition.
Int. J. of Intelligent Engineering Informatics, 2017 Vol.5, No.2, pp.167 - 190
Submission date: 21 Nov 2015
Date of acceptance: 07 Jun 2016
Available online: 08 May 2017