This chapter presents the use of multi-objective optimization for on-line automatic verification of handwritten signatures; as discriminating features of each signer are used here some functions of time and space of the position of the pen on the paper; these functions are directly used in a multi-objective optimization task in order to obtain high values of false positives indicators (FAR False Acceptance Rate) and false negatives (FFR, false rejection rate). The genetic algorithms are used to create a signer´s model that optimally characterizes him, thus rejecting the skilled forgeries and recognizing genuine signatures with large variation with respect to the training set.
TopIntroduction
The task of verifying signatures is essentially to recognize genuine signatures, rejecting imitations. The handwritten signature has many advantages for personal identification. This is the most widespread technique of identity verification for all enlightened communities (Plamondon & Lorette, 1989). It is used to give value to written trade obligations, such as promissory notes, checks, credit card vouchers, treaties, etc. Financial transactions also require the signature to be authenticated (Peerapong & Monthippa, 2010).
The handwritten signature verification is based on the assumption that the signature is the result of ballistic movements and as such, its dynamic characteristics are largely specific for a person, which are stable and repeatable. To carry the verification out is required to generate a standard reference, based on local or global features, in a direct or averaged way (Wirtz, 1997).
In the last thirty years, several research groups have worked on the design of a computer based system for automatic verification (Plamondon & Lorette, 1989). Verification techniques are classified primarily on on-line and off-line techniques. In the second case, the printed signature on a piece of paper is scanned or digitized, missing thus the information concerning to the sequence of the signature points, which is known as the dynamic of the signature. Verification off-line techniques have some advantages: only a scanner is required to obtain a digital representation of the signature, we can work with signatures on documents previously prepared and they do not need to be in digital format (Gries, 2000). Among its disadvantages is that the processing is complicated, because it requires that the signature is separated from the bottom of the document.
In the on-line verification system is used as an input device a graphic tablet (Hangai, Yamanaka, & Hamamoto, 2000) or an instrumented pen (Baron & Plamondon, 1989; Plamondon & Baron, 1989). Figure 1 shows the SmartPen pen and a digitizer tablet. The dynamic information related to the process of signing is captured. Both, dynamics data and the shape of the signature, are used for verification, which makes inherently more accurate an on-line system compared with the off-line system. A potential forger must be concerned not just on the shape (or design), but also of how the sequence and timing of certain strokes of the signature should be executed.
Figure 1. On-line signature capture Equipment: a) Instrumented Pen SmartPen, and b) digitalized tablet
The handwritten signature verification is based on the assumption that the signature is the result of ballistic movements and as such, its dynamic characteristics are largely specific for a person, which are stable and repeatable. To carry the verification out is required to generate a standard reference, based on local or global features, in a direct or averaged way (Wirtz, 1997).