A thermal comfort estimation method by wearable sensors

Abstract

Realizing an environment that could automatically modify temperature and humidity according to human thermal comfort using an air conditioning system, it is necessary to measure the quantified changes of human thermal comfort continually. While quantifying thermal comfort, human biometric data and environmental data from various environments are needed. Therefore, we proposed a method to estimate human thermal comfort through regression analysis with human biometric data acquired by wearable sensors. Since the PMV model is generally used to evaluate human thermal comfort, the PMV value as the correct thermal comfort could be calculated from the PMV formula. We constructed an experiment environment for acquiring the subject's biometric data and gained the subjects' METs and clo value for calculating the PMV value. The biometric data were analyzed in three regression models to predict PMV value, and MAE and RMSE evaluated each regression model. The fewest number of wearable sensors could be confirmed by gradually reducing the feature values of input data of the regression model. As a result, we acknowledged that the human thermal comfort in an indoor environment could be estimated by heart rate data and left-arm temperature data while applying the proposed method to a daily environment.