A novel EEG-based detection method of emergency situations for assistive vehicles

This paper presents a new Electroencephalography (EEG)-based method to detect emergency situations while drivers employ a brain-machine interface but not using limbs to operate an assistive vehicle. EEG signals were first preprocessed to remove the blinking artifact. The sums of powers of five rhythms (including alpha, delta, beta, theta, and low gamma rhythms) from 16 channels were then computed as the original feature pool. After that, Chi-square feature extraction method was employed to select features as the input of the Fisher linear classifier. The experimental results indicate that the proposed model can issue a braking command 400ms earlier than drivers with the system accuracy of 91.72% on average. The new detection model can be used to help develop a complementary driver assistant system to existing ones to improve the safety of brain-controlled driving and even driving with limbs.