Driver fatigue EEG signals detection by using robust univariate analysis

TL;DR

This paper introduces a fast, real-time method for driver fatigue detection using a single EEG channel and robust univariate analysis, achieving over 92% accuracy with minimal delay.

Contribution

It presents a novel single-channel EEG analysis technique combined with ensemble decision trees for accurate, real-time driver fatigue detection, simplifying previous multi-channel approaches.

Findings

Achieved 92.7% accuracy in fatigue detection.

Used only one EEG channel located in the left tempo-parietal region.

Detection delay of approximately 1.8 seconds.

Abstract

Driver fatigue is a major cause of traffic accidents and the electroencephalogram (EEG) is considered one of the most reliable predictors of fatigue. This paper proposes a novel, simple and fast method for driver fatigue detection that can be implemented in real-time systems by using a single-channel on the scalp. The method based on the robust univariate analysis of EEG signals is composed of two stages. First, the most significant channel from EEG raw is selected according to the maximum variance. In the second stage, this single channel will be used to detect the fatigue EEG signal by extracting four feature parameters. Two parameters estimated from the robust univariate analysis, namely mean and covariance, and two classical statistics parameters such as variance and covariance that help to tune the robust analysis. Next, an ensemble bagged decision trees classifier is used in order to discriminate fatigue signals from alert signals. The proposed algorithm is demonstrated on 24 EEG signals from the Jiangxi University of Technology database using only the most significant channel found, which is located in the left tempo-parietal region where spatial awareness and visual-spatial navigation are shared, in terms of 92.7% accuracy with 1.8 seconds of time delay.