Biometric Signature Biometric Handwritten Signature

Biometric Signature

Biometric Handwritten Signature

The use of Scale Invariant Features Transform (SIFT) measurements for the verification of handwritten signatures is increasingly well-established in the literature, and has been demonstrated as an effective means of verification and of forgery detection through empirical research (Mwangi, 2008). Ongoing research has also shown that SIFT-based algorithms for measurement and decision-making in regards to signatures and potential forgeries is more effective than the Hidden Markov Model that has been utilized more frequently in practical applications as well as in research, and advances in the specific measurements used in SIFT algorithms and mechanisms continue to increase the efficacy of this methodology while raising new research questions and potentils at the same time (Shkula & Shandilya, 2010). Further comparison and potential combination of these methods and of certain of their individual processes and parts is necessary to determine best practices in handwritten signature measurement and the creation of variance by a host of biometric factors (Shkula & Shandilya, 2010). Though substantial amounts of research into handwriting and specifically signature analysis have already been achieved in previous research, ongoing investigations in these areas are called for both by the practical community and by researchers uncertain of recent findings and established knowledge (Mwangi, 2008; Shkula & Shandilya, 2010).

Research Problem

While SIFT algorithms have proven a reliable and effective means of engaging in handwritten signature verification, there is not a clear and empirically established best system of measurement for the creation of raw data necessary in running SIFT algorithms. That is, there are multiple measurements and means of measurement that can and have been utilized with varying degrees of success in the creation of SIFT algorithms and the establishing of effective verification means, however a comparison of these means of measurement and their impact on SIFT efficacy has yet to be conducted (Shkula & Shandilya, 2010). Advances in the use of SIFT algorithms and mechanisms for handwriting analysis have both addressed and exacerbated this problem by making new recommendations that are not entirely in keeping with previous findings, leaving the full scope of SIFT efficacy and the most beneficial and accurate measurements to be used in SIFT algorithms as unknowns. This research will aim to address this gap.

Research Objectives

The primary objective of this research is to develop a SIFT-based algorithm that will improve the efficacy of handwritten signature feature extractions and enable a more powerful and accurate tool for the assessment of signature validity, with lower rates of false rejection of valid signatures and false acceptance of forged (invalid) signatures. To this end, a comparison between measurements of Euclidean distance and Mahalanobis distance in signature features will be conducted, and the most efficacious, efficient, and accurate means of measurement will be identified. From this, it will be possible to construct a more effective and accurate algorithm as the parameters of that algorithm and the data utilized in its functioning will be more accurate and more effectively and comprehensively obtained. These objectives will serve the research community in addition to providing practical benefits in signature verification and validation, reducing costs associated with discovering and failing to discover forgeries and with general handwritten signature assessment.