Multimodal Data Collection And Fusion Framework For Indoor Occupancy Prediction To Support HVAC Energy Saving

This study present a novel multimodal framework that combined data from environmental sensing and camera approaches using a parameterized classifier to validate data and employed Random Forest to predict room occupation and estimate occupancy number which addressed privacy concern and limitation of infrared sensors. This occupancy number and physiological data are then used as inputs to the proposed novel adaptive controller to manage thermal comfort and energy HVAC energy consumption.