Driving Behavior Extraction Based on Eyes Movement Patterns

Interest in autonomous driving has increased in the last decade. This generates a growing demand in developing algorithms suitable for such vehicles. In this context, modeling and understanding human driving behavior became increasingly important. The method presented in this paper addresses the problem of identifying driving behavior patterns that can be related to specific vehicle maneuvers. This method uses as input the estimation of driver gaze direction. The extracted patterns are grouped using clustering and linked with a specific driver maneuver. The validation of gaze estimation and pattern clustering was performed against two public datasets, DrivFace and DR(eye)VE.