Tracking Attention Based on EEG Spectrum

Abstract

Distraction while driving is a serious problem that can have many catastrophic consequences. Developing a countermeasure to detect the drivers’ distraction is imperative. This study measured Electroencephalography (EEG) signals from six healthy participants while they were asked to pay their full attention to a lane-keeping driving task or a math problem-solving task. The time courses of six distinct brain networks (Frontal, Central, Parietal, Occipital, Left Motor, and Right Motor) separated by Independent Component Analysis were used to build the distraction-detection model. EEG data were segmented into 400-ms epochs. Across subjects, 80% of the EEG epochs were used to train various classifiers that were tested against the remaining 20% of the data. The classification performance based on support vector machines (SVM) with a radial basis function (RBF) kernel achieved accuracy of 84.7±2.7% or 85.8±1.3% for detecting subjects’ focuses of attention to the math-solving or lane-deviation task, respectively. The high attention-detection accuracy demonstrated the feasibility of accurately detecting drivers’ attention based on the brain activities. This demonstration may lead to a practical real-time distraction-detection system for improving road safety.