Plataforma Integral Basada en IA para el monitoreo del estrés y la Optimización Climática en Call Center.

Abstract

This research focused on the non-invasive monitoring of work-related stress and fatigue in call center operators, an environment characterized by prolonged screen exposure and continuous cognitive demands. The objective was to develop an automated system capable of estimating the operator’s fatigue state and triggering environmental control actions in real time. To achieve this, a client–server architecture based on computer vision was implemented, using MediaPipe and OpenCV to extract facial landmarks and compute physiological metrics such as Eye Aspect Ratio (EAR), PERCLOS, and Mouth Ratio. These metrics were integrated into a composite indicator (StressScore) evaluated through temporal windows and managed by a backend exposing REST services and IoT integration via Home Assistant and the Broadlink RM4C Mini. The results demonstrated accurate classification of ocular and oral states, achieving 100% accuracy in detecting open and closed eyes as well as yawning events under controlled tests. The PERCLOS metric consistently differentiated normal, alert, and severe fatigue states, while the StressScore showed a coherent progression across the evaluated conditions. Finally, the environmental control module executed automatic temperature adjustments of 16°C and 18°C based on detected occupancy, validating the integration between physiological monitoring and environmental automation. Overall, the findings confirm the feasibility of the proposed system as a real-time monitoring tool aimed at improving workplace conditions without intrusive interference in the operator’s activities.