Driver identification and impostor detection based on driving behavior signals

The progressive integration of Advanced Driver Assistant Systems (ADAS) into vehicles has contributed significantly to increasing safety and comfort levels of the driver. The need to adapt the vehicle to the preferences and requirements of the driver leads to the development of individualized ADAS. Automatic identification of the driver is a key factor in the design of these systems. In this work, a driver identification model with impostor detection capability is proposed. This approach is based on non-intrusive information from driving behavior signals, and an extreme learning machine (ELM) network. The performance of the system is evaluated on the basis of groups of different number of known drivers, and possible impostor drivers. Identification rates are greater than 80% for every group category tested, and still above 90% for groups of two and three drivers. The impostor detection rate is above 80% when the car has a single genuine driver. This rate decays in inverse proportion to the number of authorized drivers, but it is greater than 50% in all cases.