Posture-Specific Spatial Calibration for Kinematic Model Optimization based on Human Activity Recognition in Ergonomic Evaluations

Abstract

Work-related musculoskeletal disorders (WMSDs) constitute the most prevalent occupational injury in manual material handling (MMH) across Europe. WMSDs lead to significant socio-economic losses. Multiple observational methods have related working conditions and ergonomics in MMH and workplaces to the risk of developing WMSDs for targeted and informed risk mitigation strategies. Digital ergonomic evaluations represent the next step in this process, complementing observational methods with statistical analysis and monitoring the evolution of risks over time. IMU-based motion capture technologies play a key role in this perspective due to their ease of integration into worker suits. However, IMU sensors are prone to estimation errors that affect the user kinematic and the task reconstruction. Optimization techniques, human activity recognition algorithm, and the posture-specific spatial calibration (PSSC) can refine the accuracy of whole-body kinematic models, enabling precise ergonomic assessments in real-time. This pilot study analyzes a standard MMH activity. Applying the proposed PSSC, the user kinematics is reconstructed with an average error of 50 mm. The optimized model improves the ergonomic risk evaluation by 60% w.r.t. the non-optimized one. Compared to the observational method, the lifting index of the digital ergonomic evaluation differs by 0.05, representing the 1.6% of the full-scale.