Simultaneous and Proportional Control Based on an Enhanced Musculoskeletal Model

Area of Science:

• Biomedical Engineering
• Neuroscience
• Rehabilitation Engineering

Background:

• Musculoskeletal models (MMs) are crucial for decoding movement intent from electromyography (EMG) signals.
• Decoding performance degrades with multi-degree-of-freedom (DoF) movements due to muscle signal crosstalk.
• Existing models struggle with complex, coordinated limb movements.

Purpose of the Study:

• To develop an enhanced musculoskeletal model (MM) for accurate 3-DoF motion prediction.
• To overcome the limitations of crosstalk in EMG signal decoding for coordinated movements.
• To improve the precision of myoelectric control for hand prosthetics.

Main Methods:

• Proposed a novel enhanced musculoskeletal model (DC-NMF-MM) integrating a "divide and conquer" (DC) strategy with non-negative matrix factorization (NMF).
• Collected surface EMG data from four muscles for wrist/MCP flexion/extension and eight channels for wrist pronation/supination from eight subjects.
• Validated the DC-NMF-MM against two control models (T-NMF-MM, P-NMF-MM) and compared with artificial neural network and linear regression models using offline and online experiments.

Main Results:

• The proposed DC-NMF-MM demonstrated superior prediction performance compared to T-NMF-MM and P-NMF-MM in offline experiments.
• DC-NMF-MM achieved higher accuracy in predicting joint angles than artificial neural network and linear regression models.
• Online experiments confirmed the feasibility of the DC-NMF-MM for effective 3-DoF motion control, showing promising results in completion time, overshoots, and path efficiency.

Conclusions:

• The enhanced DC-NMF-MM effectively addresses EMG signal crosstalk for improved 3-DoF motion prediction.
• This model offers a significant advancement for accurate and reliable myoelectric control systems.
• The developed model provides a foundation for clinical applications in hand prosthetics and rehabilitation.