Fuzzy Logic Based Driving Behavior Monitoring using Hidden Markov Models

Abstract

This paper proposes a driving behavior monitoring system to provide drivers an indicator of the danger level for driving safety. The major challenge of the research is to label whether a pattern is dangerous or not for training data. Our approach is to recognize the seven behaviors, normal driving, acceleration, deceleration, changing to the left lane or right lane, zigzag driving, and approaching the car in front by hidden Markov models. The danger-level indicator, on behalf of the dangerous situation of drivers, is inferred by fuzzy rules involved with the above behaviors. The higher the value represents the worse status of drivers, and it also provides three colors representing different levels of warnings to remind drivers of their own states. The uncertain definition of a dangerous pattern is avoided, and the related behaviors leading to the current danger level are offered instead. Moreover, unlike many studies using simulators to validate their systems and experimental results, we collected data from a real vehicle and evaluated the proposed system in a real road environment. The experimental results show that the proposed method achieved an average detection ratio of 95% for behavior recognition and can be used for driving safety.