Driver Drowsiness Detection Based on Electroencephalogram (EEG) Signal Analysis
Amy Zhang, Matt Liheng He
Port Moody Secondary, Sir Winston Churchill Secondary
Floor Location : M 114 F
Driver drowsiness has become a major issue towards ensuring safety on the road and has led to many fatal accidents, injuries, and car damages. Various studies in the past have been conducted with the purpose of detecting fatigue in drivers by examining physiological signals such as eye-closing, head nodding, and yawning. In our study, we chose to use a dry electroencephalogram (EEG) to find signals correlating to driver’s fatigue. In comparison to the other methods, we believe that the EEG is the most reliable physiological signal to make this study more effective and efficient, in addition to its low cost and easy setup. To collect data, we had a number of volunteers (mainly ourselves and family members) participate in a series of tests.
The EEG we used was a retail-level, one-channel, dry EEG machine–the Neurosky Brainwave Mobile II–which we used to identify and collect data on the different frequencies of brainwaves (Delta, Theta, Alpha, Beta and Gamma). We tested our brainwaves in different environments to collect data in varying levels of alertness, drowsiness, and sleepiness in order to identify the most predictable and prevalent brainwaves found in the drowsy stage. In collecting data, we simulated a driving environment on the computer. To improve the data’s accuracy, we developed an attention game using Java to track the exact moment when volunteers would fall asleep. Data was collected during in both alert and drowsy states.
Our data showed significant differences in our brainwaves in correlation to the stages of alertness, drowsiness, and sleepiness. We found that alpha waves had the most persistent signal during the stages of drowsiness, while other waves would only show notable features in the later stages of sleepiness. Because our project focuses on ensuring the safety of the driver, we chose to mainly use the alpha wave as the others would delay giving out a warning. A subject of interest was the absence of significant changes in the delta and theta waves as previous studies we read have had results in those areas.
Potential improvements include creating a drowsiness warning system and adapting it for use on mobile phones as a grab-and-go solution for drivers.
However, our results could be very preliminary since we did not use a multi-channel EEG that would measure other spots on the scalp which may lead to more significant changes in the brainwaves during the drowsy stages.