Improving forgery detection in off-line forensic signature processing

The handwritten signature is an established biometric modality, which finds important applications not only in typical security scenarios, but also in forensic investigations. In such situations, the amount of information available from a particular signature sample is often very limited while, on the other hand, other sources of information characteristic of an individual may in principle be exploited. This paper proposes and investigates a novel approach to individual identity prediction which integrates these diverse sources of information, encompassing both direct biometric data and additional soft biometrics. It is seen that the approach significantly enhances performance based on off-line capture of samples and also increases the flexibility with which a system can be deployed in practice.