Supporting Implicit Human-to-Vehicle Interaction: Driver Identification from Sitting Postures

Mobile internet services have started to pervade into vehicles, approaching a new generation of networked, ”smart” cars. With the evolution of in-car services, particularly with the emergence of services that are personalized to an individual driver (like road pricing, maintenance, insurance and entertainment services) the need for reliable, yet easy to handle identification and authentication has arisen. Services that demand unambiguous and unmistakable continuous identification of the driver have recently attracted many research efforts, mostly proposing video-based face/pose recognition, or acoustic analysis. A driver identification system for vehicular services is proposed, that, as opposed to video or audio based techniques, does not suffer from the continuously changing environment while driving, like lighting or noise conditions. A posture recognition technique based on a high resolution pressure sensor integrated invisibly and unobtrusively into the fabric of the driver seat has been developed, taking the pelvic bone distance as a biometric trait. Data coming from two 32x32 pressure sensor arrays (seat- and backrest) is classified according to features defined based on the pelvic bone signature, mid and high pressure distribution and body weight. Empirical studies, besides analyzing (quantitative) driver recognition performance, assess the identification technique according to the qualitative attributes universality, collectability, uniqueness, and permanency. The proposed driver identification technique is implicit and thus not reliant to attention, it is continuously in operation while seated, and requires no active person cooperation. These characteristics encourage the universal use of the approach – a whole new modality for person-to-environment interaction seems possible.