EEG-based detection of adverse mental state under multi-dimensional unsafe psychology for construction workers at height

Abstract

Working at height in construction sites is universal but dangerous, which can directly or indirectly lead to numerous injuries and fatalities. Meanwhile, workers' adverse mental state exerts a significant influence on the occurrence of safety accidents. Recent attempts have been made to precisely detect workers' unsafe psychology using electroencephalogram (EEG) technology. Unfortunately, unidimensional psychological factors considered in previous studies cannot represent complicated mental state. To fill this major knowledge gap, this study proposed a framework for comprehensively considering the effects of multi-dimensional critical unsafe psychology (i.e., fear of height, distraction, and mental fatigue) on workers’ adverse mental state at height. Results show that the four support vector machines (SVMs) achieved excellent performance with 96.33%, 96.75%, 95.50%, and 96.50% accuracy, respectively, when inputting the critical EEG features for adverse mental state assessment, verifying the effectiveness of the proposed framework. In addition, the Gaussian kernel SVM achieved 96.50% accuracy and balanced classification performance, making it most applicable to the development of adverse mental state assessment approach. The framework proposed reveals the complex interactions between unsafe psychology and adverse mental states, enriching the theoretical models of occupational safety and mental health. It provides a more comprehensive perspective on the factors influencing unsafe environments at high altitudes. This offers the possibility for the automatic detection of adverse mental states, contributing to a more proactive approach to safety management in high-altitude operations.