Signature Verification on Finger Operated Touchscreen Devices

The increasing number of smartphone users and the involved increased amount of privacy critical information that these users carry around call for improved methods to protect this data. Biometric authentication methods have one important advantage when compared to a password. A biometric trait contains more information, which is harder to copy, steal or reproduce. Modern smartphones are generally touchscreen operated and therefore suitable to acquire hand-written signatures from a user. The goal of this thesis is to implement a signature verification system that is suitable for the finger operated touchscreen devices of today. In the process different signature verification approaches are discussed and finally two separate verification systems are implemented. These are a Dynamic Time Warping based system and a global feature based system which uses a Support Vector Machine for classification. These two systems are finally combined to further increase classification performance.