Human and computer-based verification of handwritten signatures

The work reported in this thesis is set within the broad field of biometrics and, especially, is concerned with the link between human and machine -based biometric processing. The use of the handwritten signature as a biometric for authenticating the identity of individuals is investigated and aspects of the handwritten signature known to influence its susceptibility to fraudulent penetration or a false rejection of its original are particularly assessed. Two approaches to handwritten signature verification are considered: through an automatic signature verification system and by means of human visual inspection. A strong emphasis is given in the processing of static signature information. Experimental studies on human perceptual judgements regarding the complexity of signatures and their intra-class variability directed the development and assessment of automatic algorithmic solutions that reflect human perceptual criteria. A novel method is developed to quantitatively estimate the degree of a signature’s complexity based on the static signature image alone. Valuable insight is gained, through a number of experiments on human visual inspection of handwritten signatures, into strategies employed by humans in analysing signatures and the key factors that affect human performance in correctly verifying signatures’ authenticity, allowing predictions about the likelihood of a signature’s susceptibility to forgery and proposing practical scenarios for improvement. Finally, the integration of knowledge gained from the investigation of the advanced human perceptual capabilities in inspecting handwritten signature samples with automatic signature verification processing is investigated and a number of possible schemes are proposed. The possibility of improvement is, thus, offered with respect to both automatic identity authentication and also regarding human related procedures in inspecting and authenticating handwritten signature samples, which remains an important aspect of signature checking in common use.