TY - JOUR
T1 - Multimodal fusion for monitoring worker fatigue in elevated work environments
AU - Ma, Jie
AU - Li, Heng
AU - Wang, Lei
AU - Yu, Xinge
AU - Huang, Xingcan
N1 - Publisher Copyright:
© 2025 Elsevier Ltd
PY - 2025/9
Y1 - 2025/9
N2 - Falls from height are a major cause of severe accidents in the construction industry, particularly contributing to high mortality rates. Elevated construction tasks, conducted in complex and demanding environments, impose mental and physical workloads to workers, which may lead to fatigue over time. Previous studies using single-modal data to monitor worker fatigue have overlooked the multidimensional and dynamic nature of fatigue development. To address this issue, this study uses a comprehensive array of multimodal input features, including heart rate, breathing rate, skin temperature, lactate and glucose levels, and EEG signals, to classify the fatigue experienced during such tasks. For this purpose, A controlled experiment involving fifteen construction workers executing a 90-minute wall painting task on a ladder was conducted to collect data. Fatigue levels were quantified using the Borg 6–20 scale as a ground truth, with objective metrical data captured through wearable sensors. The multimodal fusion data was preprocessed prior to training random forest classification model. With an average accuracy of 94.7 %, the random forest model can effectively identify workers’ fatigue states during elevated construction tasks. This comprehensive analysis highlights the critical role of multimodal fatigue assessment in improving health and safety conditions on construction sites.
AB - Falls from height are a major cause of severe accidents in the construction industry, particularly contributing to high mortality rates. Elevated construction tasks, conducted in complex and demanding environments, impose mental and physical workloads to workers, which may lead to fatigue over time. Previous studies using single-modal data to monitor worker fatigue have overlooked the multidimensional and dynamic nature of fatigue development. To address this issue, this study uses a comprehensive array of multimodal input features, including heart rate, breathing rate, skin temperature, lactate and glucose levels, and EEG signals, to classify the fatigue experienced during such tasks. For this purpose, A controlled experiment involving fifteen construction workers executing a 90-minute wall painting task on a ladder was conducted to collect data. Fatigue levels were quantified using the Borg 6–20 scale as a ground truth, with objective metrical data captured through wearable sensors. The multimodal fusion data was preprocessed prior to training random forest classification model. With an average accuracy of 94.7 %, the random forest model can effectively identify workers’ fatigue states during elevated construction tasks. This comprehensive analysis highlights the critical role of multimodal fatigue assessment in improving health and safety conditions on construction sites.
KW - Construction worker
KW - Elevated construction tasks
KW - Fatigue
KW - Multimodal fusion
KW - Random forest model
UR - http://www.scopus.com/inward/record.url?scp=105008539136&partnerID=8YFLogxK
U2 - 10.1016/j.aei.2025.103565
DO - 10.1016/j.aei.2025.103565
M3 - 文章
AN - SCOPUS:105008539136
SN - 1474-0346
VL - 67
JO - Advanced Engineering Informatics
JF - Advanced Engineering Informatics
M1 - 103565
ER -