Miner Fatigue Detection from Electroencephalogram-Based Relative Power Spectral Topography Using Convolutional Neural...

Area of Science:

• Neuroscience
• Occupational Health
• Machine Learning

Background:

• Miner fatigue, stemming from demanding work conditions, significantly elevates safety risks and error rates in underground coal mines.
• Effective fatigue detection is crucial for preventing accidents and enhancing operational efficiency in mining environments.
• Previous research primarily employed feature-based machine learning for miner fatigue estimation.

Purpose of the Study:

• To introduce a novel method for detecting miner fatigue using electroencephalogram (EEG) signals.
• To develop a fatigue detection system capable of classifying normal, critical, and fatigue states in miners.
• To evaluate the efficacy of different feature extraction and deep learning techniques for miner fatigue detection.

Main Methods:

• Generating topographic maps from EEG signals, incorporating frequency and spatial information.
• Utilizing Power Spectral Density (PSD) and Relative Power Spectral Density (RPSD) for feature extraction.
• Applying a Convolutional Neural Network (CNN) for classifying miner states based on EEG-derived topographic maps.

Main Results:

• Relative Power Spectral Density (RPSD) demonstrated superior classification accuracy compared to Power Spectral Density (PSD) across all tested deep learning methods.
• The CNN model, when combined with RPSD features, achieved high performance metrics: 94.5% accuracy, 97.0% precision, 94.8% sensitivity, and 96.3% F1 score.
• The proposed RPSD-CNN method significantly outperformed current state-of-the-art techniques in miner fatigue detection.

Conclusions:

• The RPSD-CNN method presents a highly effective and accurate approach for detecting miner fatigue.
• This technique holds significant potential as a valuable tool for coal companies to enhance mine safety and worker efficiency.
• Further implementation of this EEG-based fatigue detection system could lead to substantial improvements in underground mining safety protocols.