A LightGBM-Based EEG Analysis Method for Driver Mental States Classification

Zeng, Hong; Yang, Chen; Zhang, Hua; Wu, Zhenhua; Zhang, Jiaming; Dai, Guojun; Babiloni, Fabio; Kong, Wanzeng · 2019 · Crossref

DOI: 10.1155/2019/3761203

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 accurately and efficiently detecting driver mental states, specifically fatigue, using electroencephalography (EEG) signals. While EEG is a promising tool for monitoring neural activity related to driving, existing methods often struggle with the signal's instability, low signal-to-noise ratio, and the high computational cost required for real-time processing. The authors propose LightFD, a lightweight classifier based on the LightGBM gradient boosting framework, designed to improve both classification accuracy and decision efficiency compared to traditional machine learning and deep learning models. The experimental data were collected from 10 healthy subjects using a driving simulation platform equipped with a 16-channel EEG amplifier. The protocol involved eight distinct stages designed to elicit specific mental states: baseline driving (WUP), performance driving (PERFO), five stages with varying audio-visual stimuli (TAV1–TAV5), and a drowsiness stage (DROWS). The researchers focused on a three-class classification problem, labeling data as awake (TAV3), neutral (TAVX, selected from TAV1, TAV2, TAV4, or TAV5 based on average track deviations), and drowsy (DROWS). EEG signals were preprocessed using band-pass filtering and independent component analysis to remove artifacts. Feature extraction utilized an improved Common Spatial Pattern (CSP) algorithm that employed channel variance for dimensionality reduction, transforming the data into a format suitable for the LightGBM model. The LightFD model incorporates a histogram-based algorithm and a leafwise tree growth strategy with depth limitation to enhance training speed and prevent overfitting. The study evaluated LightFD against Support Vector Machine (SVM), Large Margin Nearest Neighbor (LMNN), Gated Recurrent Unit (GRU), and Convolutional Neural Network (CNN) classifiers. In intrasubject classification tests, where 80% of data was used for training and 20% for testing per subject, LightFD demonstrated superior performance in terms of both accuracy and computational efficiency. The results indicated that LightFD achieved better classification metrics than the compared models while requiring significantly less time and memory resources. Additionally, the study validated LightFD’s transfer learning capabilities, showing it performed robustly in cross-subject scenarios compared to methods like Manifold Embedded Distribution Alignment (MEDA) and Metric Transfer Learning (MTLF). The findings suggest that LightFD is a highly effective solution for real-time EEG-based mental state prediction. By combining efficient feature extraction with a lightweight gradient boosting classifier, the method overcomes the computational bottlenecks associated with deep learning models and the accuracy limitations of traditional classifiers. The authors conclude that LightFD holds significant potential for practical applications in brain-computer interaction and driver monitoring systems, offering a viable path for real-time fatigue detection in vehicular environments.

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
• Theoretical Contribution: computational model