Towards wearable thermal comfort assessment framework by analysis of heart rate variability

Abstract

Buildings consume huge amounts of energy and utilize most of it for occupants’ thermal comfort satisfaction. Real-time thermal comfort assessment can enormously contribute to thermal comfort optimization and energy conservation in buildings. Existing thermal comfort models mainly utilize classification algorithms to identify personal thermal comfort states, resulting in significant information loss. In addition, existing thermal comfort studies mainly use the Fast Fourier Transform (FFT) for frequency-domain heart rate variability (HRV) feature extraction to compute the power of the Low-frequency (LF) and High-frequency (HF) bands of the R–R intervals (RRIs), resulting in the insufficient use of the information in the RRIs. To account for these concerns, this study defines personal thermal sensation as a continuous function of time, and investigates using the Hilbert Transform (HT) to extract the instantaneous amplitude (iA) of the LF and HF for thermal comfort modeling. Moreover, a novel continuous thermal sensation acquisition system has been designed to obtain the subjects’ approximate continuous thermal sensation. The FFT-based HRV features, HT-based HRV features, and time-domain HRV features have all been shown to be relevant for personal thermal comfort modeling. By utilizing machine learning regressions, it is feasible to combine the HT-based HRV features and other HRV features together to boost personal thermal comfort prediction accuracy. The subjects’ personal thermal comfort prediction reached the highest average coefficient of determination (R2) of 0.73 by using all the HRV features together. This study facilitates practical applications for wearable thermal comfort assessment frameworks and contributes to energy conservation in buildings.