Real-Time Detection and Classification of Driver Distraction Using Lateral Control Performance

This paper suggests a real-time method for detecting both visual and cognitive distraction using lateral control performance measures including standard deviation of lane position (SDLP) and steering wheel reversal rate (SRR). The proposed method adopts neural networks to construct detection models. Data for training and testing the models were collected in a driving simulator in which fifteen participants drove through a highway. They were asked to complete either visual tasks or cognitive tasks while driving to create distracted driving periods. As a result, the best performing model could detect distraction with an average accuracy of 93.1%. Keywords-driver distraction; distraction classsification; driving performance; machine learning; neural network.