A WPCA-Based Method for Detecting Fatigue Driving From EEG-Based Internet of Vehicles System

Dong, Na; Li, Yingjie; Gao, Zhongke; Ip, Wai Hung; Yung, Kai Leung · 2019 · Crossref

DOI: 10.1109/access.2019.2937914

archive: archived pipeline: cataloged verified

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

Summary

This paper addresses the critical safety issue of fatigue driving, a primary cause of traffic accidents, by proposing a real-time detection method within an Internet of Vehicles (IoV) framework. The authors identify that while electroencephalogram (EEG) signals are effective for monitoring driver fatigue, the high dimensionality of EEG data hinders real-time processing. To overcome this, the study introduces a novel Weighted Principal Component Analysis (WPCA) algorithm designed to reduce feature dimensions more effectively than traditional methods, thereby improving both detection speed and classification accuracy. The methodology involves a simulated driving experiment with eight participants using a driving simulator and a 40-channel EEG system. EEG signals were pre-processed to remove noise, and features were extracted using three distinct methods: Autoregressive (AR) models of orders 3, 4, and 5; Power Spectral Density (PSD); and Differential Entropy (DE). The core innovation, WPCA, assigns weights to features based on their individual impact on classification accuracy. Specifically, the algorithm calculates the accuracy reduction when each feature is removed, normalizes these values to create weights, and applies them to the PCA process. This ensures that features with higher discriminative power are prioritized during dimension reduction. A Support Vector Machine (SVM) served as the classifier, and the proposed WPCA-SVM approach was compared against standard SVM and PCA-SVM baselines using 10-fold cross-validation. The experimental results demonstrate that the WPCA method significantly outperforms both standard SVM and PCA-SVM across all feature extraction techniques. The 4th-order AR model yielded the best overall performance. When using WPCA-SVM with 4th-order AR features, the system achieved higher accuracy, sensitivity, and specificity compared to the other methods. Specifically, WPCA-SVM improved accuracy by approximately 4.39% over standalone SVM and by 3.11% over PCA-SVM for the 4th-order AR features. Similar improvements were observed for other feature types, with WPCA consistently providing superior classification metrics. The study confirms that weighting features according to their contribution to classification performance allows for more efficient dimension reduction without losing critical information. The significance of this work lies in its contribution to real-time IoV traffic management systems. By enhancing the efficiency and accuracy of fatigue detection, the proposed WPCA-based method enables faster processing of EEG data, which is essential for timely interventions such as warning drivers or alerting surrounding vehicles. The findings suggest that incorporating feature-weighted dimension reduction can substantially improve the reliability of physiological-based driver monitoring systems, ultimately contributing to reduced traffic accidents and improved road safety.

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