Event-based unobtrusive authentication using multi-view image sequences

Abstract

his paper presents a novel framework for dynamic activity-related user authentication utilizing dynamic and static anthropometric information. The recognition of the performed activity is based on Radon transforms that are applied on spatiotemporal motion templates. User authentication is performed exploiting the behavioural variations between different users. The upper body limb anthropometric information is extracted for each user and an attributed body-related graph structure framework is employed for the detection of static biometric features of substantial discrimination power. Finally, a quality factor based on ergonomic criteria evaluates the recognition capacity of each activity. Experimental validation illustrates that the proposed approach for integrating static anthropometric features and activity-related recognition advances significantly the authentication performance.