Detecting Thermal Discomfort of Drivers Using Physiological Sensors and Thermal Imaging

Recent technological developments have been used extensively in manufacturing vehicles in order to improve the driving experience and add multiple safety features. This article introduces a novel machine learning approach using physiological sensors and thermal imaging of the subjects to detect human thermal discomfort in order to develop a fully automated climate control system in the vehicles that does not need any explicit input from individuals. To achieve this goal, a dataset of thermal videos and physiological signals from 50 subjects is collected, an extensive analysis of different feature sets is conducted, a multimodal approach is experimented, and a cascaded classification system is proposed. Our results evidently show the capability of specific feature sets of detecting human thermal discomfort as well as the superior performance of integrating multimodal features.