EEG-Based Drowsiness Estimation for Driving Safety Using Deep Q-Learning

Fatigue is the most vital factor of road fatalities, and one manifestation fatigue during driving drowsiness. In this paper, we propose using deep Q-learning to study correlation between drowsiness performance. This carried out by analyzing an electroencephalogram (EEG) dataset captured a simulated endurance test. Driving safety research EEG data represents important brain-computer interface (BCI) paradigm from application perspective. To formulate estimation problem as optimization task, adapt terminologies in test fit reinforcement learning framework. Based on that, Q-network (DQN) tailored referring latest DQN technologies. The designed network merits characteristics can generate actions indirectly estimate results show that trained model trace variations mind state satisfactory way against testing data, which confirms feasibility practicability new computation paradigm. By comparison, it also reveals our method outperforms supervised counterpart superior for real applications. best knowledge, are first introduce BCI scenario, potentially be generalized other cases.