Personalized thermal comfort inference using RGB video images for distributed HVAC control

Highlights: • A framework of a ubiquitous thermal comfort assessment for energy-efficient HVAC. • The framework infers human thermoregulation states using RGB video images. • The framework draws on thermoregulation mechanisms and Eulerian video magnification. • Subtle blood flow variations to facial skin due to thermoregulation are inferred. • The feasibility was evaluated for 21 participants under low and high temperatures. HVAC systems account for more than 40% of energy consumption in buildings to provide satisfactory indoor environments for occupants. The integration of personalized thermal comfort in the operation of HVAC systems has been shown to be highly effective in enhancing energy efficiency of buildings. To this end, research efforts have proposed personalized thermal comfort assessment through voting (i.e., occupant feedback) and profiling as well as physiological response measurement. In this study, we have proposed a novel approach for enabling RGB video cameras as sensors for measuring personalized thermoregulation states – an indicator of thermal comfort. If their feasibility for thermoregulation state inference could be established, optical cameras provide a cost-effective and omnipresent solution for distributed measurement of thermal comfort and consequently control of HVAC systems for energy saving. Accordingly, we have proposed a framework that draws on the concepts of thermoregulation mechanisms in the human body as well as the Eulerian video magnification approach. The framework is composed of several components including face detection, skin pixels isolation, image magnification. And calculation of detection index to infer subtle blood flow variations to the facial skin surface (i.e., blood perfusion), which is due to thermoregulation adjustments. In order to minimize the impact of variable illumination condition and the ambient noise on the results, different combinations of methods for framework components were taken into account. The feasibility assessments were conducted through an experimental study with 21 participants under low (20 °C) and high (30 °C) temperatures. In total, 16 positive cases out of 18 statistically significant cases were observed resulting in 89% of success rate using the most promising combinations of the methods. The results demonstrate that the proposed framework could contribute to realization of a non-intrusive, cost-effective, and ubiquitous distributed thermal comfort assessment that has been proven critical in increasing energy efficiency of the HVAC system through distributed control feedback.