Investigating real-time thermal sensation predictions using predicted mean vote and local skin temperature analysis via infrared thermography as a nonintrusive tool

Predicting thermal sensation without individual query is challenging. The predicted mean vote (PMV) is a widely used mathematical model for estimating thermal sensation vote (TSV). However, its accuracy relies on precise evaluation of environmental factors, metabolic rate (M), and clothing insulation (Icl). Furthermore, M and Icl are difficult to measure and often assumed to be uniform, leading to prediction errors. Similarly, physiological factors such as skin temperature (tsk) and proximal-to-distal skin temperature gradient (Delta tproximal-distal), which correlate with TSV, are challenging to measure directly. This study explored the use of infrared thermography (IRT) as a nonintrusive method for PMV prediction, with tsk and Delta tproximal-distal as determinants of TSV. Experiments were conducted during summer, autumn, and winter in the Hokkaido University, Japan. Icl was estimated using (1) questionnaire surveys, (2) clothing temperature (tcl), and (3) a combination of tcl with tsk, both derived from IRT. M was estimated using (1) a standard uniform value (1.2 MET), (2) heart rate (HR), and (3) estimated basal metabolic rate (BMR) combined with physical activity ratio (PAR). Local tsk and Delta tproximal-distal were evaluated at the face and hand regions using IRT. Results indicated that PMV predictions based on M from BMR and PAR, and Icl derived from tcl and tsk, correlated more strongly with TSV compared to traditional methods using uniform values. Additionally, specific skin temperatures (tnose, thand) and gradients (Delta tforehead-nose, Delta teye canthus-nose) were significantly associated with TSV. These findings demonstrate that IRT is an effective nonintrusive tool for real-time thermal sensation prediction.