Prediction of Individual Dynamic Thermal Sensation in Subway Commute Using Smart Face Mask.

Area of Science:

• Environmental Science
• Biomedical Engineering
• Machine Learning

Background:

• Wearable sensors and machine learning are common for predicting thermal sensation.
• Existing studies often lack real-world applicability due to controlled lab settings and inconvenient sensors.
• Dynamic ambient conditions are crucial but often overlooked in thermal sensation prediction.

Purpose of the Study:

• To predict individual dynamic thermal sensation using physiological and psychological data.
• To develop and validate a smart face mask for real-time thermal sensation monitoring.
• To assess the correlation between physiological data and perceived thermal comfort in natural environments.

Main Methods:

• Designed a smart face mask measuring skin temperature (SKT) and exhaled breath temperature (EBT).
• Conducted real-time human experiments in a subway cabin with 20 male students under natural conditions.
• Utilized a smartphone application for data collection and wavelet decomposition for feature engineering.
• Employed the bagged tree algorithm for individual thermal sensation prediction.

Main Results:

• The bagged tree model achieved high accuracy (98.14%) and F1-score (96.33%) for predicting thermal sensation.
• Individual thermal sensation showed significant correlations with measured SKT, EBT, and derived features.
• The smart face mask effectively captured physiological data in dynamic, real-world settings.

Conclusions:

• The developed smart face mask and machine learning model accurately predict individual dynamic thermal sensation.
• Physiological data, particularly SKT and EBT, are strong indicators of thermal comfort in natural environments.
• This approach offers a practical solution for personalized thermal comfort assessment in real-world scenarios.