Thermal Adaptation Behavior Recognition Based on Video Understanding

The individual differences in thermal environment perception are an important reason why the current actual building environment is insufficient for people to reach the expected thermal satisfaction. Personnel thermal state recognition technology and personalized environmental adjustment systems are generally regarded as key technologies with application potential. By analyzing behavioral characteristics (such as rubbing hands, wiping sweat, adjusting clothing, etc.), we can better understand an individual's response to the thermal environment. In this study, video data is first used to capture an individual's behavior patterns under high temperature or cold conditions, and then behavior classification is performed based on spatiotemporal feature extraction and deep neural network models. Experiments show that the proposed model has high accuracy in recognizing different thermal adaptation behaviors, and the final accuracy reaches <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{8 8. 9 \%}$</tex>, indicating the feasibility of RGB video data in inferring thermal sensation states and providing theory and technology for the development of a new generation of artificial intelligence-driven real-time thermal sensation recognition technology.