Dynamic modeling of EMG-force relationship using parallel cascade identification.

Area of Science:

• Biomechanics and Motor Control
• Biomedical Engineering
• Signal Processing

Background:

• Surface electromyography (sEMG) is a non-invasive method to record muscle electrical activity.
• Accurate estimation of joint torques and forces from sEMG is crucial for understanding motor control and developing advanced prosthetics.
• Existing methods often struggle to capture the complex, nonlinear relationship between muscle activation and joint force.

Purpose of the Study:

• To introduce and evaluate Parallel Cascade Identification (PCI) as a novel dynamic estimation tool.
• To map upper-arm muscle sEMG recordings to elbow-induced wrist force.
• To compare the performance of PCI against established methods like orthogonalization schemes.

Main Methods:

• PCI utilizes a parallel structure combining linear dynamic and nonlinear static blocks.
• sEMG data from upper-arm muscles were recorded.
• The PCI model was trained and tested to predict wrist force.
• PCI performance was experimentally compared to an orthogonalization scheme.

Main Results:

• PCI demonstrated superior performance in predicting wrist-induced force compared to the orthogonalization scheme.
• The enhanced predictive capability of PCI is attributed to its ability to model nonlinear dynamic effects.
• PCI provides a more accurate mapping from muscle activity to joint force.

Conclusions:

• Parallel Cascade Identification (PCI) is an effective tool for dynamic force estimation from sEMG.
• PCI's architecture is well-suited for capturing nonlinear dynamics in biological systems.
• This approach offers significant improvements for applications requiring accurate force prediction from muscle signals.