A non-invasive facial multi-region temperature recognition and comfort prediction

Real-time understanding of human thermal comfort state is important for efficient HVAC system control and building energy efficiency. This paper proposes a non-invasive face multi-region temperature recognition and thermal comfort prediction method. The temperature features of multiple regions on human face are identified by infrared camera, and six temperature sensitive regions are determined by applying sensitivity analysis; Volunteers are recruited for facial temperature collection using infrared thermography and subjective thermal sensation voting (TSV); Cheeks and nose are identified as the most sensitive regions for thermal comfort by correlation analysis; Nose and cheek temperatures are used as inputs to predict human thermal comfort state online by using support vector machine (SVM). Results show that accuracy of the 3-category thermal states achieves to 93.87%.