A Novel Experiment Design for Vision-Based Fatigue Detection

Abstract

Manufacturing companies continuously integrate robots for collaboration with human workers. That challenges the design of a safe and ergonomic worker’s place to avoid collisions with the robot or other possible accidents that can severely injure a human. As many tasks in manufacturing premises entail repetitive tasks like carrying heavy loads back and forth over a long period, one critical aspect that the industry focuses on is the detection of human fatigue state. While the literature targets bio-markers like the arousal state or measures deviations in the walking or carrying pattern using inertial measurement units (IMU) or other body sensors, studies considering a vision-based approach are sparse. Additionally, the usage of specific body devices demands individual calibration and is prone to errors in the sensor readings. Therefore, we introduce and explain in detail a novel experimental protocol for fatigue induction in humans performing a bucket load-carry task. The experimental design considers only a camera setup (RGB and neuromorphic) that records the face and posture of each participant, delivering data applicable for feature extraction and the development of a fatigue detection module at a later stage. Finally, we provide some preliminary evaluation from the pilot study obtained from the Swedish Occupancy Fatigue Inventory (SOFI) questionnaire.