Abstract:
Scientific literature on biometric identity identification and verification dates back to 18th
century. However, the development of automated biometric authentication and verification
systems is not that much old as it advances with the advancement in computer processing which
itself is an emerging area. From the available biometric authentication systems and methods,
Handwritten Signature Verification (HSV) systems have taken over as the most emerging and
reliable means of authentication/verification. The rising reliance on electronic storage and
transmission of documents has raised a need for an online means of electronically verifying the
identity of sender/author. This research presents an efficient and robust Online Signature
Verification (OSV) system targeting verification rates better than the available state-of-the-art
systems in the presence of skilled forgeries. Fourier analysis is employed on the signatures
followed by Linear Fisher Discriminant Analysis (LFDA) to obtain compressed (to get lower
dimensional) representation while enhancing inter-class scatter between signature patterns.
Signature modeling is performed using m-mediod-based modeling approach where m-mediods
are put on to represent data distribution in each class. Our mediod-based model is generated in
three steps. First, we tend to model the distribution data for each class with m representative
mediods upper-bounded by the # of class samples. In the second step, we tend to identify
conceivable normality ranges for our mediod-based model through various system parameters
tuning. Finally, normality ranges for each identified mediod are identified by exploiting the data
distribution around that mediod. Euclidean Distance (ED) is used as dis/similarity. A total of
1560 signature samples including skilled forgeries are considered in our study. The evaluation of the proposed system on Japanese signature dataset provided by SigWiComp2013 realized
promising results than the competitors.
