Abstract
The detection of driver distractions is exceptionally important for driving safety. Driver distraction can originate from various sources such as external tasks (e.g., texting or eating) or mental states (e.g., sleepiness, tiredness, anger, and tension). To detect these conditions, most of the previous studies were based on vision-based techniques. These techniques are affected by environmental factors (e.g., day, night, and facial accessories such as glasses and hats). However, the steering wheel is an interface that provides a direct relationship between the driver and vehicle. The driver’s interaction can effectively reflect this behavior and mental state. This study introduced a new method for detecting driver distractions by utilizing force/torque (F/T) sensor data extracted from the steering wheel. An experimental setup was designed and developed to measure the accuracy of the proposed method. To validate the strategy, a machine learning-based algorithm was developed. It demonstrated remarkable performance in determining the position of the driver’s hand on the steering wheel and in inferring with high precision the hand the driver uses to operate the vehicle. The method produced accurate results in all the grip ranges that could be held by the driver within the range of 0°–360°. The support vector machine (SVM) method was used in machine learning. It predicted with a 91.1% accuracy rate.