Multiplex Limited Penetrable Horizontal Visibility Graph from EEG Signals for Driver Fatigue Detection

Cai, Qing; Gao, Zhong-Ke; Yang, Yu-Xuan; Dang, Wei-Dong; Grebogi, Celso · 2019 · Crossref

DOI: 10.1142/s0129065718500570

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 challenge of detecting driver fatigue, a significant contributor to road accidents, by investigating the underlying brain mechanisms through electroencephalography (EEG) signals. While various methods exist for monitoring driver behavior or physiological signals, the characterization of the neural mechanisms associated with mental fatigue remains insufficiently understood. To bridge this gap, the authors propose a novel method called the Multiplex Limited Penetrable Horizontal Visibility Graph (Multiplex LPHVG). This approach transforms multichannel EEG time series into complex brain networks, allowing for both the detection of fatigue states and the analysis of topological changes in brain connectivity. The experimental design involved ten right-handed students performing a simulated driving task lasting approximately 90 minutes in a monotonous environment designed to induce fatigue. EEG signals were recorded using a 40-channel cap at a sampling frequency of 1000 Hz. Data preprocessing included downsampling to 250 Hz, band-pass filtering (1–50 Hz), and Independent Component Analysis to remove eye-blink artifacts. The study defined "alert" states using the first 5 minutes of the session and "fatigue" states using the last 5 minutes. The Multiplex LPHVG method constructs a multilayer network where each layer corresponds to an EEG channel. The method calculates the average edge overlap to measure overall coherence and quantifies interlayer correlations to build a weighted brain network. Network topology was characterized using weighted global efficiency, weighted clustering coefficient, and weighted characteristic path length across a range of sparsity levels. The results demonstrate that the Multiplex LPHVG method effectively distinguishes between alert and fatigue states with high accuracy using a Support Vector Machine classifier combined with leave-one-out cross-validation. Statistical analysis via paired t-tests revealed significant differences in network measures between the two states. Specifically, the study found a significant increase in the weighted clustering coefficient as the brain transitioned from an alert state to a mental fatigue state. This finding suggests that mental fatigue is associated with increased local connectivity or segregation in the brain network. The integration of average edge overlap with traditional network metrics provided a robust feature set for classification, validating the efficacy of the proposed method. The significance of this work lies in its dual contribution to both practical fatigue detection and theoretical understanding of brain dynamics. By successfully applying complex network theory to multivariate EEG data, the study provides a computationally efficient and analytically tractable tool for monitoring driver mental states. Furthermore, the observed increase in clustering coefficient offers novel insights into the cognitive processes and neural reorganization associated with mental fatigue, moving beyond simple detection to characterize the specific brain behaviors underlying driver impairment.

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
• drowsiness
• microsleep

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