From Wearable Sensor Networks to Markerless Motion Capture for Instrumental-Based Biomechanical Risk Assessment in...

Area of Science:

• Ergonomics and Occupational Biomechanics
• Human-Computer Interaction
• Wearable Technology

Background:

• Manual material handling is a primary cause of work-related low-back disorders.
• Accurate biomechanical risk assessment is crucial for effective prevention strategies.
• Existing wearable sensor networks have limitations in ergonomics assessments.

Purpose of the Study:

• To compare markerless (ML) motion capture with wearable sensor networks for evaluating lifting tasks.
• To assess the accuracy and consistency of ML systems in calculating biomechanical risk metrics (RWL, LI).
• To explore the potential of ML for training AI algorithms in automatic biomechanical risk classification.

Main Methods:

• Twenty-eight workers performed standardized lifting tasks under varying risk conditions.
• Kinematic variables were captured using both wearable sensor networks and a multi-camera ML system.
• Data were analyzed to compute Recommended Weight Limit (RWL) and Lifting Index (LI) based on the revised NIOSH equation.

Main Results:

• ML systems demonstrated closer agreement with reference benchmarks and lower variability compared to wearable sensors, except at low risk levels (LI=1).
• Significant differences were observed between the two systems for most kinematic measures.
• ML-based kinematic data achieved comparable accuracy to wearable systems for automatic biomechanical risk classification.

Conclusions:

• Markerless motion capture technology shows significant potential for accurate, repeatable, and cost-effective biomechanical risk assessment in occupational ergonomics.
• ML systems offer a promising alternative to wearable sensors for evaluating demanding lifting tasks.
• ML approaches can provide valuable input for developing AI-driven tools for automated ergonomic risk assessment.