A NOVEL METHOD FOR DRIVER INATTENTION DETECTION USING DRIVER OPERATION SIGNALS

摘要

This paper presents a new methodology for the analysis of driver inattention using the driving operation signals. The proposed method constructs a driver model based on a nonlinear autoregressive exogenous input (NARX) model using neutral driving data. The nonlinear function of the model is approximated by a multilayer perceptron (MLP) neural network. Driving data with and without a secondary task, collected using a real vehicle in Nagoya city, was used in this study. The performance of the model was evaluated during a validation process where the model produced acceptably small errors and the predicted values closely following the actual values. Finally, through an assessment of the model residuals between the actual signal and predicted signal of the inattentive driving data, we prove that driver inattention while driving in a real environment can be detected.