Detecting subject-specific fatigue-related changes in lifting kinematics using a machine learning approach.

Area of Science:

• Biomechanics
• Ergonomics
• Machine Learning

Background:

• Fatigue responses during lifting tasks are highly individual, necessitating subject-specific detection methods.
• Current fatigue assessment often relies on subjective measures, lacking objectivity in workplace settings.
• One-class support vector machines (OCSVM) offer a potential objective approach for classifying kinematic changes associated with fatigue.

Purpose of the Study:

• To evaluate the efficacy of a subject-specific OCSVM approach for objective fatigue detection during repetitive lifting.
• To determine if changes in lifting kinematics, identified as outliers by OCSVM, correlate with self-reported fatigue levels.

Main Methods:

• Participants performed a repetitive lifting protocol while motion capture recorded kinematic data.
• Subject-specific OCSVM models were trained using the initial lifting motions (first 35%) of each participant.
• Subsequent lifts were classified as outliers against the trained OCSVM decision boundaries, and the percentage of outliers was correlated with the rating of perceived exertion (RPE).

Main Results:

• A significant positive correlation was observed between the percentage of outlier lifts and RPE in participants who exhibited fatigue.
• No significant correlation was found for participants who did not report increased fatigue.
• The OCSVM successfully identified changes in lifting kinematics indicative of fatigue.

Conclusions:

• An OCSVM-based outlier detection tool shows prospective efficacy for subject-specific, objective fatigue detection in repetitive lifting.
• This machine learning approach can identify changes in movement patterns associated with increased self-reported fatigue.
• The findings support the development of objective fatigue monitoring systems for occupational safety and performance optimization.