Human movement analysis as a measure for fatigue: a hidden Markov-based approach

Area of Science:

• Biomechanics
• Machine Learning
• Sports Science

Background:

• Exercise performance is significantly altered by fatigue, affecting movement kinematics.
• Monitoring fatigue during rehabilitation and sports is crucial for injury prevention.
• Existing methods for fatigue estimation may lack precision in capturing dynamic changes.

Purpose of the Study:

• To investigate the efficacy of a parametric hidden Markov model (PHMM) for estimating exercise-induced fatigue.
• To compare the performance of PHMM against linear regression in fatigue detection.
• To develop a model that incorporates exercise progress and subject-specific initial conditions.

Main Methods:

• Utilized a parametric hidden Markov model (PHMM) to analyze kinematic data from exercise movements.
• Developed a top-level hidden Markov model with variable state transitions to represent fatigue progression.
• Tested the PHMM approach on a squat exercise database recorded using optical motion capture.
• Compared PHMM performance against linear regression models.

Main Results:

• The PHMM demonstrated a high correlation between estimated fatigue and subjective fatigue ratings.
• Accurate fatigue estimation was achieved for single squats, sets of squats, and entire exercise routines.
• The PHMM approach showed superior performance compared to linear regression in capturing fatigue dynamics.

Conclusions:

• Parametric hidden Markov models offer a robust and accurate method for estimating exercise-induced fatigue from kinematic data.
• This approach can enhance the monitoring of fatigue in both rehabilitation and sports settings.
• The developed PHMM provides a valuable tool for objective fatigue assessment, potentially reducing injury risk.