Event Related Biometrics: Towards an Unobtrusive Sensing Seat System for Continuous Human Authentication

摘要

The present work is focused on the improvement of a Sensing Seat system previously developed by the authors for the initial authentication purpose in office and car scenarios. The goal is to obtain an event-related continuous authentication system, where the human subject should not take care of the system itself so that he is free to perform his normal actions. The system is realized by means of a sensing cover where conductive elastomers are used as strain sensors. The deformation of the cover caused by the body shape while actions are performed by the subject are used to obtain time-dependent relevant features. Such information are then analyzed by suitable classifiers that are able to perform the real-time continuous authentication task. A measurement campaign was carried out using data from 24 human subjects employed in an office scenario while a set of 22 actions were performed. The authentication capabilities of the system are reported in terms of acceptance and rejection rates, showing a high degree of correct classification.