Predicting Method of Simple Reaction Time of Driver Based on SVR Model

In order to accurately predict the simple reaction time of drivers to emergencies and construct the basis of adaptive warning system in dangerous driving state, the forecasting method of the driver's simple reaction time based on Electroencephalogram (EEG) in real time was proposed. First, characteristic parameters were extracted from the EEG by the fast Fourier transform (FFT) as objective forecasting indexes of drivers'simple reaction time. The forecasting model of the simple reaction time of drivers to emergencies based on the support vector regression (SVR) was established. Finally, 4 hours continual driving reaction time and EEG from 20 drivers were used to test the model. The results indicates that three EEG characteristic parameters (θ, α, β) are significantly correlated with the simple reaction time and the correlation of EEG characteristic parameter α is the most significant, which provides the objective prediction index for predicting the reaction time of the SVR model. The SVR model is constructed by the kernel function in terms of radial basis function (RBF), Polynomial and Sigmoid function and the simple reaction time is predicted as well. The various errors of SVR based on the RBF function are lower than those in consideration of two other functions, which indicates that the prediction precision of SVR model using RBF kernel function is the best, and the accuracy of prediction is above 80%. © 2017, Editorial Department of China Journal of Highway and Transport. All right reserved.