A Flexible Architecture for Online Signature Verification Based on a Novel Biometric Pen

In this article, a flexible online signature verification architecture is presented. Signatures are recorded with a novel pen, the so-called biometric smart pen (BiSP). The dynamic of a signature is captured by sensors mounted inside the pen. Thus, no specific tablet is needed. The recorded signature signals (i.e., signature time series) are presented to the authentication system, where they are classified by different static and dynamic classifiers using different features (i.e., parametric and functional features) extracted in a preceding feature extraction stage. In order to get a single decision whether a signature is genuine or forged, the individual results of every classifier are combined in a final decision stage using an ensemble technique. A key feature of the presented system is the possibility to configure the whole reference model for each person individually. Almost every stage of the proposed architecture (segmentation, preprocessing, feature extraction and selection, classification and decision) can be configured in a person-specific way. Experiments show that our flexibly configurable system provides a reliable authentication with an accuracy of 99.6%