A quick identification model for assessing human anxiety and thermal comfort based on physiological signals in a hot and humid working environment

We propose a simple and quick method for quantifying workers' anxiety and thermal comfort levels using physiological signals. Nine subjects enrolled in a series of controlled laboratory experiments involving varying temperature, relative humidity, and labor intensities. A total of 40 experiments were conducted, and 1592 groups of anxiety data and 1624 groups of thermal comfort data were obtained, respectively. During 2-h-working trials, Electroencephalogram (EEG), photoplethysmography (PPG), and pupil diameter of each subject were collected synchronously, and the State-Trait Anxiety Inventory (STAI) and thermal comfort vote (TCV) were completed in stages. Random Forest was adopted to screen out the appropriate sensitivity feature indicators of anxiety levels and thermal comfort levels from the 70 features of the 10 EEG channels. Finally, Random Forest, Gradient Boosting Decision Tree, K-nearest Neighbor Algorithm, and Support Vector Machine were used to determine relevant physiological data combinations and modeling algorithms. The Precision of the anxiety level and thermal comfort level quick identification model based on Random Forest Algorithm can reach 81.04% and 84.79%, respectively. This suggests that the proposed quick identification method for assessing workers' anxiety and thermal comfort levels holds promise. Physiological data need to be obtained by monitoring only PPG, pupil diameter, and 5 EEG channels. By processing these data, the workers' anxiety and thermal comfort level could be judged realistically to ensure their safety. It is suggested that PPG, pupil diameter, and EEG should be considered all together in the future study of anxiety and thermal comfort.