From Facial Expressions to Thermal Sensation: POMS-Validated AI-Based Mood Estimation Driving Psychology-Adaptive HVAC Control

Abstract

The psychological state of building occupants plays a critical role in how thermal environments are perceived and experienced, yet conventional Heating, Ventilation, and Air Conditioning (HVAC) control strategies largely ignore this factor. Current systems typically focus on environmental and personal parameters, overlooking the influence of mood state on thermal comfort. This omission can lead to suboptimal comfort levels, decreased occupant satisfaction, and inefficient energy use. Integrating psychological feedback into the HVAC control has the potential to transform indoor climate management into a truly occupant-centric process. To this aim, this study presents a novel framework that employs artificial intelligence (AI) and image processing together to estimate occupants' mood states in real time. Facial expressions are analysed using a deep learning-based computer vision model, and the resulting mood predictions are validated with the Profile of Mood States (POMS) questionnaire to ensure accuracy and reliability. Validated mood data directly informs the HVAC setpoint adjustments, enabling psychology-adaptive control that responds dynamically to occupants' current mood states. Moreover, the system operates in real time, combining low-latency image analysis with adaptive control algorithms to continuously align thermal conditions with validated mood estimations. Additionally, implementing mood-driven HVAC control shows potential for enhancing perceived comfort while improving energy efficiency. By bridging the gap between psychological state assessment and environmental control, this research contributes to the advancement of intelligent building systems, paving the way for more responsive, energy-conscious, and human-centered indoor environments.