Error-Rate Based Biometrics Fusion

摘要

This paper addresses the face verification problem by fusing visual and infra-red face verification systems. Unlike the conventional least squares error minimization approach which involves fitting of a learning model to data density and then perform a threshold process for error counting, this work directly formulates the required target error count rate in terms of design model parameters. A simple power series model is adopted as the fusion classifier and our experiments show promising results.