Design of a Sensorized Safety Helmet for Stress and Fatigue Assessment via Real-Time Oculometric Activity

Industrial workplaces that require continuous concentration and physical effort often see accidents triggered by elevated cognitive workload, mental fatigue, and stress. To detect these early precursors, the present thesis introduces a sensorized safety helmet that combines various physiological and behavioural sensors into a single, wearable device. The helmet incorporates electroencephalography (EEG) for neural activity, electrodermal activity (EDA) for arousal variations, photoplethysmography (PPG) for cardiovascular response, an inertial measurement unit (IMU) for head posture and sudden movements, and a camera for oculometric measures. All modules are contained in lightweight customized cases mounted in a commercially available industrial helmet, while a Raspberry Pi 5 processor acquires every signal and runs real-time warnings. Although the complete system values each sensing channel, the research behind this thesis focuses on oculometric measures. The camera, positioned under the helmet in a way that does not obstruct the wearer’s view, captures three features: blink rate, gaze direction, and eye aspect ratio (EAR). These measures provide the earliest evidence of decreasing alertness and were therefore chosen as the primary indicators for the study. The physical prototype and related software were tested in a laboratory that reproduces a logistics warehouse, where volunteers carried out picking tasks whose difficulty increased in well-defined steps. Throughout each session the software logged every signal continuously and stamped a warning whenever a threshold was exceeded, for example when head inclination persisted, blink dynamics became irregular, or the gaze wandered outside the work zone. Each event was time-stamped for later inspection, allowing brief lapses in attention to be highlighted before they could escalate into safety-critical incidents. Beyond the real-time monitoring, an offline classification model was built to explore how the recorded signals relate to perceived cognitive workload. After each trial, participants completed the Task Load Index form developed by the National Aeronautics and Space Administration (NASA), a validated subjective workload assessment tool. The physiological features—primarily those derived from eye tracking—were compared with the questionnaire scores to identify patterns linked to stress, mental and physical fatigue. Early results show that reduced gaze stability, increased blink frequency, and lower EAR values correspond with higher self-reported workload, particularly in the attention and mental demand dimensions. By combining immediate warnings with post-task workload analysis, the helmet delivers a two-tier strategy: immediate detection of risk conditions and broader insight into human workload performance during industrial tasks. These findings confirm the promise of integrating oculometrics with complementary physiological measures in standard safety equipment, with the goal of enhancing adaptive safety systems in cognitively and physically demanding environments. Future work will reinforce the robustness of the model, validate the system in active warehouse operations, and investigate advanced methods for merging the different sensor streams into a single fatigue-risk index.