Classifying lower extremity muscle fatigue during walking using machine learning and inertial sensors

Area of Science:

• Biomechanics
• Machine Learning
• Human Movement Analysis

Background:

• Lower extremity muscle fatigue impairs postural stability, motor performance, and normal walking.
• Automated fatigue detection can aid in early identification of fall and injury risks.
• Support vector machines (SVMs) are effective for classifying gait patterns.

Purpose of the Study:

• To explore the potential of SVMs in recognizing gait patterns associated with lower extremity muscle fatigue.
• To classify fatigue and non-fatigue conditions using an inertial measurement unit (IMU).
• To investigate the influence of different SVM kernel schemes on classification accuracy.

Main Methods:

• Recorded kinematic and kinetic gait patterns of 17 participants during normal and fatigued walking.
• Induced lower extremity fatigue through squatting exercises to 60% maximal voluntary exertion.
• Utilized feature selection methods on temporal and frequency domain signals for SVM classification.
• Compared linear, polynomial, and radial basis function (RBF) kernels for SVM.

Main Results:

• Lower extremity muscle fatigue significantly altered gait and loading responses.
• Achieved 96% accuracy in distinguishing fatigued from non-fatigued gait patterns within subjects.
• Linear and RBF kernels demonstrated comparable effectiveness in identifying intra-individual fatigue characteristics.

Conclusions:

• Intra-subject fatigue classification using gait patterns from IMUs shows significant potential.
• This approach can help identify individuals at risk of falls or injuries due to muscle fatigue.
• SVMs provide a robust tool for automated gait-based fatigue assessment.