Real-Time Task Discrimination for Myoelectric Control Employing Task-Specific Muscle Synergies.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Robotics

Background:

• Myoelectric control of lower arm prostheses remains a significant challenge.
• Existing methods often struggle with real-time accuracy and controllability.
• Understanding muscle configurations is key to improving prosthetic function.

Purpose of the Study:

• To develop a novel, robust, and computationally efficient framework for myoelectric control.
• To enhance the discrimination accuracy and real-time performance of prosthetic limb control.
• To leverage muscle synergy hypothesis and state-space representation for improved prosthetic function.

Main Methods:

• Utilized task-specific muscle synergies and state-space representation of neural signals.
• Modeled synergy activation coefficients as latent system states estimated via a constrained Kalman filter.
• Employed a post-processing algorithm with posterior probabilities for task discrimination from electromyogram (EMG) data.

Main Results:

• Achieved >90% discrimination accuracy in approximately 3 milliseconds.
• Demonstrated superior performance compared to common machine learning algorithms in both off-line and real-time evaluations.
• Showcased high robustness and computational efficiency in real-time prosthetic control.

Conclusions:

• The proposed formulation offers a significant advancement in myoelectric control for lower arm prostheses.
• The algorithm provides accurate and efficient real-time control, enhancing prosthetic usability.
• This approach holds promise for developing more intuitive and responsive prosthetic devices.