Measuring Biomechanical Risk in Lifting Load Tasks Through Wearable System and Machine-Learning Approach

Area of Science:

• Occupational Health
• Biomechanics
• Wearable Technology

Background:

• Work-related musculoskeletal disorders are a significant cause of non-fatal occupational injuries.
• Correct posture is crucial for minimizing stress on the back and lower extremities during manual material handling.
• Real-time ergonomic evaluations using biomechanical parameters offer potential for injury reduction.

Purpose of the Study:

• To propose a system for recognizing postural patterns using wearable sensors and machine learning.
• To evaluate the effectiveness of kinematic data in differentiating between correct and incorrect postures.
• To assess the impact of load weight on lifting kinematics.

Main Methods:

• Twenty-six healthy subjects performed manual material handling tasks with correct and incorrect postures.
• Eight wireless inertial measurement units (IMUs) collected kinematic data.
• A biomechanical model estimated joint range of motion and trunk displacement.
• A support vector machine (SVM) was trained for posture classification.

Main Results:

• Significant statistical differences (p < 0.01) were observed in all kinematic parameters between correct and incorrect postures.
• Increased load weight led to significant changes (p < 0.01) in hip and trunk kinematics during lifting.
• The SVM achieved 99.4% accuracy (100% specificity) using all kinematic parameters for posture recognition.

Conclusions:

• Wearable IMUs combined with machine learning provide a highly accurate method for real-time ergonomic posture assessment.
• This approach can effectively distinguish between correct and incorrect postures, aiding in the prevention of musculoskeletal injuries.
• Kinematic data, particularly trunk segment parameters, are vital for developing effective automated ergonomic evaluation systems.