Cross-Subject Zero Calibration Driver's Drowsiness Detection: Exploring Spatiotemporal Image Encoding of EEG Signals...

Area of Science:

• Neuroscience
• Computer Science
• Automotive Engineering

Background:

• Road accidents are a significant cause of fatalities, with driver fatigue being a major contributing factor.
• Drowsiness detection is crucial for advanced driver assistance systems (ADAS) to mitigate risks associated with fatigue.
• Electroencephalogram (EEG) signals are valuable for recognizing mental states but face challenges like low signal-to-noise ratio and cross-subject variability.

Purpose of the Study:

• To explore and evaluate two distinct methodologies for drowsiness detection using EEG signals.
• To focus on cross-subject zero calibration, reducing the need for individual subject-specific training.
• To assess the effectiveness of these methods in a sustained-attention driving task.

Main Methods:

• Utilized electroencephalogram (EEG) signals from a public dataset of 27 subjects.
• Employed spatiotemporal image encoding representations of EEG signals: recurrence plots and Gramian angular fields.
• Applied deep convolutional neural networks (CNNs) for classification of drowsiness states.

Main Results:

• Achieved a superior balanced accuracy of up to 75.87% using leave-one-out cross-validation.
• Demonstrated comparable performance between recurrence plots and Gramian angular fields for drowsiness detection.
• Showcased the feasibility of cross-subject zero calibration, a significant advancement for practical applications.

Conclusions:

• The proposed EEG-based drowsiness detection methods, utilizing image encoding and CNNs, show significant promise.
• The successful implementation of cross-subject zero calibration paves the way for more accessible and practical ADAS.
• These findings contribute to enhancing road safety by enabling reliable fatigue detection in drivers.