Towards Unsupervised Incremental and Proportional Myocontrol Based on Higher-Density Surface Electromyography

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Technology

Background:

• Upper limb differences significantly impact daily autonomy.
• High abandonment rates of current prostheses stem from complexity and limited functionality.
• Advanced myocontrol strategies are crucial for enhancing prosthetic capabilities.

Purpose of the Study:

• To investigate fully unsupervised incremental myocontrol using higher-density surface electromyography (sEMG).
• To assess the feasibility of real-time extraction of muscle synergies via incremental sparse non-negative matrix factorization (ISNMF).
• To evaluate user performance and workload with an increasing number of controllable synergies.

Main Methods:

• Employed two 32-channel sEMG bracelets for real-time muscle synergy extraction.
• Utilized incremental sparse non-negative matrix factorization (ISNMF) for unsupervised learning.
• Tested eight able-bodied participants using a virtual target achievement control (TAC) test with varying synergy numbers.

Main Results:

• Participants demonstrated independent control of up to six muscle synergies in full-intensity tasks.
• Proportional control proved challenging, with a median success rate of 10% for half-intensity targets.
• Subjective workload showed minimal increase despite enhanced task complexity.

Conclusions:

• Fully unsupervised myocontrol using ISNMF shows promise for advanced prosthetic applications.
• Further refinement of training protocols and hyperparameter tuning is necessary.
• Validation on users with limb differences is required to confirm clinical utility.