User authentication on mobile devices with dynamical selection of biometric techniques for optimal performance

Biometrics-based authentication exhibits unique advantages in preventing stolen, lost, or forgotten identification, and therefore is being increasingly used on handheld and mobile devices for access control and electronic transactions. Aimed at developing a new method to deal with the practical issue of uncontrolled noise conditions in mobile computing, this paper applies the principle of Design of Experiment for systematic performance evaluation of biometric techniques, and proposes an authentication system capable of dynamically selecting combinations of normalization and fusion methods for optimal performance in different working environments. Experiment results are also included in the paper with analysis and discussions.