A ROBUST FRAMEWORK FOR DRIVER FATIGUE DETECTION FROM EEG SIGNALS USING ENHANCEMENT OF MODIFIED Z-SCORE AND MULTIPLE MACHINE LEARNING ARCHITECTURES

Abdubrani, Rafiuddin; Mustafa, Mahfuzah; Zahari, Zarith Liyana · 2023 · Crossref

DOI: 10.31436/iiumej.v24i2.2799

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 fatigue, a major contributor to traffic fatalities, by proposing a robust framework for detecting fatigue states using electroencephalogram (EEG) signals. The research is motivated by the challenges inherent in analyzing portable EEG data, which often contains significant artifacts such as muscle activity and eye blinks. To overcome these limitations and improve classification reliability, the authors developed a methodology that integrates signal preprocessing, enhanced feature extraction, and multiple machine learning architectures. The experimental design utilized EEG data from 12 healthy male participants aged 19 to 24, recorded during a driving simulation task. Data was collected from eight specific channels (O1, O2, Fp1, Fp2, P3, P4, F3, and F4) using a Neuroscan device at a sampling rate of 1000 Hz. The study distinguished between normal and fatigued states, with fatigue self-reported by participants after 40–100 minutes of driving. Preprocessing involved Independent Component Analysis (ICA) to remove unwanted noise and artifacts. For feature extraction, the researchers employed a Continuous Wavelet Transform (CWT) using the Morlet mother wavelet to analyze alpha, delta, and theta sub-bands. A key innovation was the application of an enhanced modified z-score, which uses the median and median absolute deviation rather than mean and standard deviation, to robustly handle outliers and enhance statistical features. The framework evaluated five classifiers: Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Artificial Neural Networks (ANNs) with Long Short-Term Memory (LSTM), and Support Vector Machines (SVMs). Performance was assessed using 5-fold cross-validation. The results demonstrated exceptional precision, with the framework achieving an average accuracy of 96.07%. The ANN classifier yielded the highest individual accuracy at 99.65%, followed by the SVM at 97.89%. Receiver Operating Characteristic (ROC) analysis confirmed outstanding performance across all classifiers, with an average Area Under the Curve (AUC) value of 0.95. The significance of this work lies in its comprehensive approach to improving the accuracy and robustness of EEG-based driver fatigue detection. By combining multichannel EEG analysis with an enhanced modified z-score for outlier removal, the study addresses common data quality issues that hinder reliable classification. The high accuracy achieved by multiple machine learning models suggests that this framework is a viable candidate for real-time driver monitoring systems, potentially contributing to reduced accident rates by providing early warnings of fatigue.

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
• Methodological Resource: tool software