Neural Networks on Handwritten Signature Verification

Neural Networks on Handwritten Signature Verification

J. Francisco Vargas (University of Las Palmas de Gran Canaria, Spain & Universidad de Antioquia, Colombia) and Miguel A. Ferrer (University of Las Palmas de Gran Canaria, Spain)
Copyright: © 2009 |Pages: 6
DOI: 10.4018/978-1-59904-849-9.ch180
OnDemand PDF Download:


Biometric offers potential for automatic personal identification and verification, differently from other means for personal verification; biometric means are not based on the possession of anything (as cards) or the knowledge of some information (as passwords). There is considerable interest in biometric authentication based on automatic signature verification (ASV) systems because ASV has demonstrated to be superior to many other biometric authentication techniques e.g. finger prints or retinal patterns, which are reliable but much more intrusive and expensive. An ASV system is a system capable of efficiently addressing the task of make a decision whether a signature is genuine or forger. Numerous pattern recognition methods have been applied to signature verification. Among the methods that have been proposed for pattern recognition on ASV, two broad categories can be identified: memory-based and parameter-based methods as a neural network. The Major approaches to ASV systems are the template matching approach, spectrum approach, spectrum analysis approach, neural networks approach, cognitive approach and fractal approach. The proposed article reviews ASV techniques corresponding with approaches that have so far been proposed in the literature. An attempt is made to describe important techniques especially those involving ANNs and assess their performance based on published literature. The paper also discusses possible future areas for research using ASV.
Chapter Preview


As any human production, handwriting is subject to many variations from very diverse origins: Historic, geographic, ethnic, social, psychological, etc (Bouletreau, 1998). ASV is a difficult problem because signature samples from the same person are similar but not identical. In addition, a person signature often changes radically during their lifetime (Hou, 2004). Although these factors can affect a given instance of a person writing, writing style develops as the writer learns to write, as do consistencies which are typically retained (Guo, 1997). One of the methods used by expert document examiners is to try to exploit these consistencies and identify ones which are both stable and difficult to imitate. In general, ASV systems can be categorized into two kinds: The On-line and Off-line systems. For On-line, the use of electronic devices to capture dynamics from signature permits to register more information about the signing process while improving the system performance, in the case of Off-line approaches for ASV, this dynamic information is lost and only a static image is available. This makes it quit difficult to define effective global or local features for the verification purpose.

Three different types of forgeries are usually take into account on ASV system: random forgeries, produced without knowing either the name of the signer nor the shape of his signature; simple forgeries, produced knowing the name of the signer but without having an example of his signature; and skilled forgeries, produced by people who, looking at an original instance of the signature, attempt to imitate it as closely as possible. The problem of signature verification become more difficult when passing from random to simple and skilled forgeries, the later being so difficult a task that even human beings make errors in several cases. It is pointing out that several systems proposed up to now, while performing reasonably well on a single category of forgeries, decrease in performance when working with all the categories simultaneously, and generally this decrement is bigger than one would expect.(Abuhaiba,2007; Ferrer,2005).

Key Terms in this Chapter

Backpropagation algorithm: Learning algorithm of ANNs, based on minimising the error obtained from the comparison between the outputs that the network gives after the application of a set of network inputs and the outputs it should give (the desired outputs).

Agent Based Mode: A specific individual based computational model for computer simulation extensively related to the theme in complex systems, Monte Carlo Method, multi agent systems, and evolutionary programming. The idea is to construct the computational devices (agents with some properties) and then, simulate them in parallel to model the real phenomena.

Feature Selection: The technique, commonly used in machine learning, of selecting a subset of relevant features for building robust learning models. Its objective is three-fold: improving the prediction performance of the predictors, providing faster and more cost-effective predictors, and providing a better understanding of the underlying process that generated the data.

Genetic Algorithms: A genetic algorithm is technique used for searching or programming. It is used in computing to find true or approximate solutions to optimization and search problems of various types and used as a function in evolutionary computation. Genetic algorithms are based on biological events. They mimic biological evolution.

Automatic Signature Verification: A procedure that determine if a handwritten signature is genuine or a forgery, when a person claims for identity verification.

Principal Component Analysis: A technique used to reduce multidimensional data sets to lower dimensions for analysis. PCA involves the computation of the eigenvalue decomposition of a data set, usually after mean centering the data for each attribute.

Fuzzy Logic: Derived from fuzzy set theory dealing with reasoning that is approximate rather than precisely deduced from classical predicate logic. It can be thought of as the application side of fuzzy set theory dealing with well thought out real world expert values for a complex problem.

Complete Chapter List

Search this Book: