[Paper Review] Hearables: Ear EEG Based Driver Fatigue Detection

Metin Yarıcı, Pierluigi Vito Amadori+5 more|arXiv (Cornell University)|Jan 16, 2023

Sleep and Work-Related FatiguePsychology3 citations

TL;DR

This study demonstrates that ear-EEG, using wearable hearables, can detect driver mental fatigue with comparable accuracy (70% accuracy, MCC 0.4) to conventional scalp EEG, leveraging machine learning on delta, theta, alpha, and beta band power features. It provides the first empirical evidence that ear-EEG captures fatigue-related EEG changes consistent with scalp EEG, enabling real-time, ultra-wearable fatigue monitoring.

ABSTRACT

Ear EEG based driver fatigue monitoring systems have the potential to provide a seamless, efficient, and feasibly deployable alternative to existing scalp EEG based systems, which are often cumbersome and impractical. However, the feasibility of detecting the relevant delta, theta, alpha, and beta band EEG activity through the ear EEG is yet to be investigated. Through measurements of scalp and ear EEG on ten subjects during a simulated, monotonous driving experiment, this study provides statistical analysis of characteristic ear EEG changes that are associated with the transition from alert to mentally fatigued states, and subsequent testing of a machine learning based automatic fatigue detection model. Novel numerical evidence is provided to support the feasibility of detection of mental fatigue with ear EEG that is in agreement with widely reported scalp EEG findings. This study paves the way for the development of ultra-wearable and readily deployable hearables based driver fatigue monitoring systems.

Motivation & Objective

• To investigate the feasibility of detecting driver mental fatigue using ear-EEG as a less obtrusive alternative to scalp EEG.
• To compare ear-EEG performance against conventional scalp EEG in detecting fatigue-related EEG band changes.
• To develop and evaluate a machine learning model for automatic fatigue classification using only physiologically meaningful EEG features from ear-EEG.

Proposed method

• Conducted a simulated monotonous driving task with ten subjects wearing both ear-EEG and scalp EEG sensors.
• Extracted fourteen EEG features per channel: mean power and peak frequency for delta, theta, alpha, and beta bands (excluding beta due to EMG noise).
• Used a gradient-boosted XGBoost classifier with five-fold cross-validation to assess fatigue detection performance.
• Evaluated model performance using accuracy and Matthews Correlation Coefficient (MCC), with MCC preferred for balanced binary classification.
• Analyzed feature importance to identify which EEG features contributed most to fatigue prediction.
• Compared single-channel ear-EEG performance to standard scalp EEG channels (Fz, Cz, POz) for model benchmarking.

Experimental results

Research questions

• RQ1 Can ear-EEG detect the same fatigue-related EEG band changes (delta, theta, alpha) as scalp EEG during monotonous driving?
• RQ2 How does the performance of a machine learning model trained on ear-EEG compare to one trained on scalp EEG for fatigue classification?
• RQ3 Which EEG features (mean power vs. peak frequency) are most informative for fatigue detection in ear-EEG?
• RQ4 Is the ear-EEG signal sufficiently robust and representative of brain activity to enable real-time fatigue monitoring?
• RQ5 Can a single-ear EEG channel achieve comparable fatigue detection performance to standard scalp EEG channels?

Key findings

• Ear-EEG achieved a fatigue classification accuracy of 70% and MCC of 0.4, comparable to a three-channel scalp EEG setup (74% accuracy, MCC 0.48).
• The top-performing features in the ear-EEG model were mean power features across frequency bands, indicating that power levels are more informative than peak frequency for fatigue detection.
• Feature importance analysis showed that multiple mean power features contributed significantly to predictions, suggesting a rich information content in ear-EEG signals.
• Single-ear EEG performance was on par with standard scalp EEG channels: Fz (69% acc, MCC 0.39), Cz (68% acc, MCC 0.68), POz (70% acc, MCC 0.4).
• The study provides novel numerical evidence that ear-EEG reflects the same key EEG changes—particularly increased theta and alpha power—associated with mental fatigue as seen in scalp EEG.
• The model demonstrated real-time feasibility using 10-second measurement windows, supporting deployment in practical, continuous monitoring systems.