Detecting slow eye movement for recognizing driver's sleep onset period with EEG features

Area of Science:

• Neuroscience
• Sleep Research
• Human-Computer Interaction

Background:

• Slow eye movements (SEM) are established indicators of sleep onset period (SOP).
• The specific characteristics and utility of SEM for detecting driving fatigue remain understudied.
• Previous research has not fully explored the relationship between SEM, eye closure events (ECEs), and electroencephalography (EEG) during driving simulations.

Purpose of the Study:

• To investigate the characteristics of SEM during eye closure events (ECEs) in simulated driving.
• To evaluate the potential of SEM as a reliable indicator for recognizing a driver's sleep onset period (SOP).
• To develop and validate an improved algorithm for SEM detection by incorporating EEG features.

Main Methods:

• Visual observation of SEM during ECEs in ten subjects' experimental data.
• Analysis of ECE duration distribution using box plots to assess sleepiness levels.
• Development of a novel SEM detection algorithm combining horizontal electrooculogram (HEOG) signals with EEG power features from the occipital O2 signal.
• Feature selection using the maximum relevance and minimum redundancy (mRMR) method.
• Classification of SEM using a support vector machine (SVM).

Main Results:

• SEMs were observed to occur predominantly during eye closure events (ECEs).
• ECEs with SEM, particularly those accompanied by alpha wave attenuation, exhibited longer durations, indicating higher sleepiness levels.
• The proposed algorithm integrating EEG power features improved SEM detection accuracy by an average of 1.4% compared to using HEOG signals alone.
• The EEG feature P(α+θ)/β was identified as the most significant predictor for SEM detection.

Conclusions:

• SEMs are reliable indicators of a driver's sleep onset period (SOP), especially when associated with alpha wave attenuation during ECEs.
• Incorporating EEG power features significantly enhances the accuracy of SEM detection algorithms.
• The findings support the development of advanced systems for monitoring driver fatigue and sleepiness.