Fatigue is one of the causes of falling asleep at the wheel, which can result in fatal accidents. Thus, it is necessary to have practical fatigue detection solutions for drivers. In literature, electroencephalography (EEG) along with the surrogate measure of reaction time (RT) has been used to develop fatigue detection algorithms. However, these solutions are often based upon wet multi-channel EEG electrodes which are not feasible or practical for drivers. Using dry electrodes and headband like designs would be better. Hence, this study aims to investigate the correlation of EEG log bandpower against RT via a Muse headband which has dry frontal EEG electrodes. 31 subjects underwent an hour-long driving simulation experiment with car deviation events. Based on the video and EEG data, 5 'Sleepy' and 5 'Alert' subjects are identified and analyzed. A differential signal between Fpl and Fp2 is computed so as to remove the effects of eye blinks, and is analyzed for correlation with RT. Significant positive correlation is found for log delta (1-4 Hz) bandpower, and significant negative correlations for log theta (4-8 Hz) and alpha (8-12 Hz) bandpowers, but the positive correlation of log beta (12-30 Hz) bandpower with RT is not significant. This is a good first step towards building a practical fatigue detection solution for drivers in the future.
展开▼
