An Electromyographic-driven Musculoskeletal Torque Model using Neuro-Fuzzy System Identification: A Case Study.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Technology

Background:

• Estimating joint torque from electromyography (EMG) is crucial for understanding muscle function and developing assistive devices.
• Previous models often lack clinical interpretability or sufficient accuracy in dynamic conditions.

Purpose of the Study:

• To develop and validate a novel neuro-fuzzy model for estimating elbow joint torque from high-density surface EMG signals.
• To enhance the clinical interpretability and reduce estimation error in EMG-Torque modeling.

Main Methods:

• Utilized high-density surface EMG signals from biceps brachii and triceps brachii muscles during isometric elbow flexion-extension.
• Applied principal component analysis (PCA) for EMG amplitude estimation.
• Developed a neuro-fuzzy model to establish the relationship between EMG amplitudes and measured torque, optimizing model complexity.

Main Results:

• The neuro-fuzzy model achieved high accuracy, with a % variance accounted for of 96.40 ± 3.38%.
• The optimal model complexity was consistently found to be 4 ± 1 rules, suggesting underlying motor control strategies.
• Demonstrated improved clinical interpretability compared to traditional black-box models and reduced estimation error versus state-of-the-art nonlinear models.

Conclusions:

• The proposed neuro-fuzzy approach offers a promising, interpretable, and accurate tool for EMG-Torque modeling.
• This method has significant potential for clinical applications in movement analysis and rehabilitation engineering.