Construction workers are frequently exposed to biomechanical risk factors in terms of muscle activity, which contribute to work-related musculoskeletal disorders. Current assessment methods often rely on sensor-based motion capture systems, which can be costly and time-consuming. Therefore, this study examines highly accurate musculoskeletal analysis that does not require motion sensors by conducting and comparing analysis using motion sensors and AI-based motion capture systems. Twelve male university students participated in this study. Inertial measurement units (IMUs) and a video recording camera were used to collect the data. An AI-based (Plask) motion capture system and a musculoskeletal model were used to analyze the data. Four lower back muscles (i.e., the erector spinae, multifidus, psoas major, and quadratus lumborum) were examined for biomechanical analysis. Results showed that the highest muscle activity was exhibited by the erector spinae muscle at 0 cm height for both IMUs and Plask motion capture system during the tasks. The Root Mean Square Errors (RMSE) for both tasks were very small (<6%), suggesting that the Plask (AI-based) motion capture systems can evaluate biomechanical risk with minimal errors. Therefore, the AI-based motion capture (Plask) system is recommended for future biomechanical risk factor assessments using sensors-less data collection methods.
Yazaki et al. (Wed,) studied this question.