Object Classification in the Fixation of a Car Driver

Driver state detection is a key topic in the complex area of automated driving. The reaction time of the driver is critical in the context of partial and conditional automated driving. Hence, it is important to know what is on the driver’s mind. To gain this knowledge, a head mounted eye tracking (ET) device is used to determine the drivers fixation. This paper presents a method for automatic processing of the eye tracking data with neural networks in order to automatically classify fixated objects inside and outside the car. Further, a comparison between various structures of the eye tracking signal processing shows differences in precision and calculation time. This approach reduces the effort for the analysis of human subject studies involving eye tracking. Therefore, a higher number of participants and objects could be processed in faster time. In addition, this method can be used to evaluate other unobtrusive methods for driver attention detection.