Wireless ear EEG to monitor drowsiness

Kaveh, Ryan; Schwendeman, Carolyn; Pu, Leslie; Arias, Ana C.; Muller, Rikky · 2024 · Crossref

DOI: 10.1038/s41467-024-48682-7

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the critical safety issue of drowsiness-induced accidents, which contribute significantly to fatal vehicle crashes and industrial hazards. While existing monitoring solutions like camera-based eye tracking or steering sensors have limitations regarding obstructions and false alarms, electrophysiological (ExG) monitoring offers higher accuracy. However, current neural wearables often require bulky electronics, wet electrodes needing skin preparation, or user-specific calibration, hindering their adoption for everyday, long-term use. The authors aim to demonstrate the feasibility of a wireless, dry-electrode, in-ear EEG system that is user-generic, comfortable, and capable of accurate drowsiness detection without extensive setup. The researchers developed a modular, 3D-printed earpiece using a clear methacrylate photopolymer, designed in small, medium, and large sizes to fit diverse demographics. The device features four in-ear electrodes positioned within the first 10 mm of the ear canal and two out-ear reference electrodes. To ensure low impedance and durability without hydrogels, the team employed a novel electroless plating process to create dry, gold-plated electrodes on a copper-nickel substrate. This fabrication method, reminiscent of printed-circuit-board techniques, allows for rapid prototyping and extended electrode lifespan. The earpieces were integrated with existing wireless, high-bandwidth neural recording hardware (WANDmini) to transmit data to a base station. Data were collected from nine subjects over 35 hours while they performed drowsiness-inducing tasks involving reaction time measurements and Likert scale self-reports. Drowsy events were labeled based on thresholds derived from these behavioral metrics. The study extracted temporal and spectral features from the EEG and electrooculography (EOG) signals. Three classifier models were trained using user-specific, leave-one-trial-out, and leave-one-user-out cross-validation splits. The support vector machine (SVM) classifier achieved the highest performance, reaching an accuracy of 93.2% when evaluating known users and 93.3% when evaluating never-before-seen users. These results indicate that the system can generalize effectively across different individuals without user-specific training. The findings demonstrate that wireless, dry-electrode in-ear devices can classify drowsiness with accuracy comparable to state-of-the-art wet-electrode scalp and in-ear systems. The study highlights the viability of population-trained classification models for future electrophysiological applications, reducing the need for individual calibration. By combining additive manufacturing for user-generic earpieces with robust, low-maintenance dry electrodes and offline classical machine learning algorithms, this work establishes a foundation for discreet, long-term, and scalable neural monitoring systems suitable for real-world environments such as driving and industrial operations.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• drowsiness detection algorithms

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: physiological data