The most commonly used protection mechanisms today are based on either what a person possesses (e.g. an ID card) or what the person remembers (like passwords and PIN numbers). However, there is always a risk of passwords being cracked by unauthenticated users and ID cards being stolen, in addition to shortcomings like forgotten passwords and lost ID cards (Huang & Yan, 1997). To avoid such inconveniences, one may opt for the new methodology of Biometrics, which though expensive will be almost infallible as it uses some unique physiological and/or behavioral (Huang & Yan, 1997) characteristics possessed by an individual for identity verification. Examples include signature, iris, face, and fingerprint recognition based systems. The most widespread and legally accepted biometric among the ones mentioned, especially in the monetary transactions related identity verification areas is carried out through handwritten signatures, which belong to behavioral biometrics (Huang & Yan,1997). This technique, referred to as signature verification, can be classified into two broad categories - online and off-line. While online deals with both static (for example: number of black pixels, length and height of the signature) and dynamic features (such as acceleration and velocity of signing, pen tilt, pressure applied) for verification, the latter extracts and utilizes only the static features (Ramesh and Murty, 1999). Consequently, online is much more efficient in terms of accuracy of detection as well as time than off-line. But, since online methods are quite expensive to implement, and also because many other applications still require the use of off-line verification methods, the latter, though less effective, is still used in many institutions.
In general, all the current off-line signature verification systems can be divided into the following sub-modules:
Preprocessing and Noise Removal
Feature Extraction and Parameter Calculations
Learning and Verification (or Identification)