Smart IoT-driven biosensors for EEG-based driving fatigue detection: A CNN-XGBoost model enhancing healthcare quality

Rezaee, Khosro; Nazerian, Asmar; Ghayoumi Zadeh, Hossein; Attar, Hani; Khosravi, Mohamadreza; Kanan, Mohammad · 2024 · Crossref

DOI: 10.34172/bi.30586

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 driving fatigue, a leading cause of traffic accidents, particularly in extreme environments where traditional detection methods often fail due to harsh climatic conditions and physiological variability. The authors propose a novel Internet of Medical Things (IoMT) framework that integrates smart biosensors with an advanced artificial intelligence model to detect driver fatigue in real-time. The primary motivation is to overcome the limitations of existing systems, such as high computational complexity, reliance on handcrafted features, and insufficient accuracy in cross-individual scenarios, thereby enhancing healthcare quality and road safety. The methodology employs a hybrid machine learning approach termed CNN-XGBoost Evolutionary Learning. Electroencephalogram (EEG) signals were collected from 16 participants (aged 17–25) using a 32-electrode brain helmet during driving simulation tests designed to replicate real-world conditions. To handle the non-stationary nature of EEG data, the researchers applied Discrete Wavelet Transform (DWT) for frequency decomposition and Continuous Wavelet Transform (CWT) combined with a Hanning window to convert time-domain signals into RGB scalogram images. These images were processed by a 2D Convolutional Neural Network (2DCNN) to extract spatial and temporal features, which were then fed into an Extreme Gradient Boosting (XGBoost) classifier. The model’s parameters were optimized using a modified Particle Swarm Optimization (PSO) algorithm to enhance feature selection and classification performance. The proposed CNN-XGBoost model achieved a remarkable accuracy of 99.80% on the driver fatigue dataset, significantly surpassing the performance of existing methods reviewed in the literature, which ranged from approximately 79% to 99.65%. The integration of CWT-based scalogram generation allowed for effective feature extraction from complex EEG signals, while the fusion of CNN and XGBoost leveraged the strengths of deep learning for feature learning and boosting algorithms for robust classification. The system demonstrated high reliability and reduced computational load, making it suitable for real-time applications in resource-constrained edge devices. The significance of this work lies in its successful establishment of an AIoT infrastructure for critical driving conditions. By optimizing data processing and reducing computational complexity, the system enables accurate and timely fatigue detection in extreme environments, such as mining areas or polar regions. This approach not only improves the precision of fatigue monitoring but also offers a scalable solution for integrating biosensors into broader healthcare and safety networks, potentially reducing road fatalities and enhancing the quality of life in smart societies.

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