Using Pre-stimulus EEG to Predict Driver Reaction Time To Road Events

Rahman, Shams Ur; O'Connor, Noel; Lemley, Joe; Healy, Graham · 2022 · Crossref

DOI: 10.1109/embc48229.2022.9870904

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 driver reaction times to road events, which are often impaired by mental fatigue and drowsiness. While existing methods for monitoring driver state, such as psychometric tests, are intrusive, passive sensing techniques like Electroencephalography (EEG) offer a non-intrusive alternative. The primary research question is whether pre-stimulus EEG data can predict a driver’s reaction time to specific road events using subject-independent models. This approach is motivated by the practical need for real-world systems that do not require individual calibration for each driver, thereby enabling broader deployment in driver safety assistance systems. The researchers utilized a publicly available dataset containing 32-channel EEG recordings from 27 subjects performing a sustained-attention driving task in a simulated four-lane highway environment. After discarding three subjects due to anomalous data, the study focused on 24 participants. The experimental design involved inducing random lane deviation events, with reaction time defined as the interval between the deviation onset and the driver’s corrective response offset. The core methodology involved extracting 2-second EEG epochs immediately preceding each event. Power Spectral Density (PSD) features were computed for delta, theta, alpha, and beta frequency bands. To ensure generalizability, the authors employed a leave-one-subject-out cross-validation strategy, training models on 23 subjects and testing on the remaining subject. Two machine learning algorithms were evaluated: Bayesian Ridge Regression and Artificial Neural Networks (ANN), with a dummy regressor serving as the baseline. The results demonstrated that both machine learning models outperformed the baseline dummy regressor across all frequency bands. Specifically, features from the alpha and theta bands yielded the lowest Mean Absolute Error (MAE), consistent with literature linking these bands to attention and mental fatigue. The ANN model generally achieved slightly lower errors than Bayesian Ridge Regression. On a per-trial basis, the models successfully predicted individual reaction times for the majority of subjects, with statistically significant correlations between predicted and ground truth values. Furthermore, when aggregating predictions per subject, the models significantly correlated (Pearson-r = 0.71, p = 0.00008) with the subjects' average reaction times. This indicates that the system could effectively classify drivers as fast or slow responders based on their pre-event neural activity. The significance of this work lies in its demonstration that subject-independent EEG models can reliably predict driver reaction times prior to the occurrence of road events. By validating the utility of alpha and theta band features in a cross-subject framework, the study provides a viable pathway for developing real-time driver monitoring systems. These systems could potentially trigger autonomous interventions when a driver’s neural state suggests an impending sub-optimal reaction, thereby enhancing road safety without the logistical burden of per-user model training. Future work aims to optimize channel selection and explore deep learning architectures to further improve prediction accuracy.

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
• anticipation

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, behavioral performance data
• Theoretical Contribution: computational model