Real-Time and Dynamically Consistent Estimation of Muscle Forces Using a Moving Horizon EMG-Marker Tracking...

Area of Science:

• Biomechanics and Motor Control
• Computational Physiology
• Rehabilitation Engineering

Background:

• Accurate muscle force estimation is crucial for effective clinical biofeedback and movement analysis.
• Existing methods like static optimization have limitations, including unrealistic joint torques and inconsistent estimates during co-contraction.
• Forward optimal control approaches offer dynamic consistency but are computationally expensive for real-time applications.

Purpose of the Study:

• To assess the feasibility and accuracy of real-time muscle force estimation using a Moving Horizon Estimation (MHE) approach.
• To evaluate the computational efficiency and performance of MHE compared to existing methods.

Main Methods:

• A 4-Degrees of Freedom (DoFs) arm model with 19 Hill-type muscles was developed.
• Simulations included varying levels of co-contraction, electromyography (EMG) noise, and marker noise.
• MHE was implemented with different cost functions (EMG-marker tracking, marker-tracking with muscle excitation minimization) and tested on excitation- and activation-driven models.

Main Results:

• An excitation-driven model with a 7-frame MHE achieved real-time performance (24 Hz), a 3,500-fold speed improvement over prior methods.
• Estimation errors for muscle forces ranged from 1-30 N (EMG-marker tracking) and 8-50 N (muscle excitation minimization) under simulated noise.
• Statistical analysis confirmed significant effects of co-contraction and noise levels on estimation accuracy.

Conclusions:

• The MHE implementation demonstrates significant potential for accurate, real-time muscle force estimation in biofeedback applications.
• This approach overcomes computational limitations of previous methods, offering a viable tool for clinical movement analysis.
• The MHE method shows promise even under realistic experimental noise conditions.