A practical method for driver sleepiness detection by processing the EEG signals stimulated with external flickering light

Abstract

A novel steady state visual evoked potential (SSVEP)-based BCI system for driver’s sleepiness monitoring is proposed. Detecting the driver’s concentration is one of the most challenging assignments for researchers in these decades. A continuous attention and awareness is necessary to drive safely. Otherwise, one tragedy of drowsy driving would probably happen. Therefore, real-time sleepiness detection can restrain accidents effectively. In this study, SSVEPs are used for running the proposed system and two experimental setups consisting four single and paired light-emitting diodes (LEDs) using two different fast Fourier transform-based feature extraction methods, and three different classifiers of the linear discriminant analysis, the support vector machine (SVM) and the Max ones on the accuracy of the system are studied. For real-time application, related features are extracted from four different sweep lengths (temporal durations) of 0.5, 1, 2, and 3 s. The experimental results show that higher sweeps have higher accuracies and the SVM classifier, experimental setup of 4-paired LEDs in sweep length of 3 s has the highest accuracy of 98.2 %, while with the comparable information transfer rate (ITR) value of 24 bits/min within the sweep length of 1 s, this time is considered as the best response time. Therefore, this study demonstrates the feasibility of the proposed system in a practical driving application.