Prediction on Thermal Sensation and Thermal Comfort from Multi-Dimensional Environmental and Physiological Characteristics

Thermal sensation and thermal comfort are subjective feelings of human, which are affected by physiological, psychological and environmental factors. Most of the traditional universal models and personalized models focus on the prediction of human thermal sensation and comfort using the single characteristic, such as the heart rate or the environment temperature. However, these studies lack comprehensive consideration for environmental and physiological factors. In this study, we use a deep learning framework by combining both multi-dimensional physiological and environmental characteristics for thermal sensation and comfort prediction. In particular, considering the time dependency of input characteristics, causal convolution neural network is embedded in the framework. Results show that the prediction accuracy of the universal model based on datasets consisted of 72 subjects is 32.53% for the thermal sensation and 30.68% for the thermal comfort. The accuracy of the personalized model based on the dataset including one subject is 99.00% for the thermal sensation and 90.00% for the thermal comfort. These results show that the prediction on the thermal sensation and comfort from multi-dimensional environmental and physiological characteristics is feasible. It further indicates that the multi-dimensional characteristics have the application potential for the future researches of the prediction of human subjective feelings.