Isokinetic elbow joint torques estimation from surface EMG and joint kinematic data: using an artificial neural...

Area of Science:

• Biomechanics
• Neuroscience
• Artificial Intelligence

Background:

• The relationship between electromyographic (EMG) signals and isometric joint torque is not well understood.
• Predicting joint torque from muscle activity is crucial for understanding human movement and developing assistive technologies.

Purpose of the Study:

• To determine the relationship between EMG activity and isokinetic elbow joint torque using an artificial neural network (ANN) model.
• To validate the predictive accuracy of the ANN model with experimental data.

Main Methods:

• A 3-layer feed-forward artificial neural network (ANN) was developed using an error back-propagation algorithm.
• The model was trained and tested using rectified, low-pass filtered EMG signals, joint angle, joint angular velocity, and measured torque.
• Sensitivity analysis was performed on the number of hidden nodes.

Main Results:

• The ANN model accurately predicted isokinetic joint torque from novel EMG activities, joint position, and angular velocity.
• Model predictions showed high correlation with experimental data (correlation coefficient gamma = 0.998 in learning, 0.900 in testing).
• Electrode placement significantly impacted the ANN model's accuracy.

Conclusions:

• Artificial neural networks can effectively model the complex relationship between EMG signals and joint torque in human isokinetic movements.
• The developed ANN model provides a reliable method for predicting joint torque from muscle activity.
• Optimizing electrode site selection is critical for improving the accuracy of EMG-based torque prediction models.