Consistent control information driven musculoskeletal model for multiday myoelectric control.

Area of Science:

• Biomedical Engineering
• Human-Machine Interaction
• Rehabilitation Robotics

Background:

• Electromyography (EMG)-driven musculoskeletal models (MMs) are crucial for human-machine interaction.
• Non-stationary EMG signal characteristics degrade the long-term performance of EMG-driven MMs.
• Consistent muscle excitation extraction is needed for reliable, long-term control.

Purpose of the Study:

• To develop a robust, zero-retraining muscle excitation extraction approach for EMG-driven MMs.
• To improve the estimation performance of MMs for simultaneous hand and wrist movement prediction.
• To address the non-stationary nature of EMG signals in long-term myoelectric control.

Main Methods:

• An improved non-negative matrix factorization (NMF) algorithm was developed by adding constraints and L2-norm regularization terms.
• The enhanced NMF was applied to extract stable muscle synergies and their time-varying profiles.
• The extracted muscle synergies were used to drive a musculoskeletal model for movement prediction.

Main Results:

• The proposed method demonstrated significantly superior and robust performance in inter-day experiments compared to existing methods.
• Competitive methods included machine learning, EMG envelope-driven MMs, and classic NMF-based MMs.
• Analysis confirmed the effectiveness of the method in achieving consistent muscle excitations across different days.

Conclusions:

• The developed NMF-based approach offers a promising pathway for robust, zero-retraining control of myoelectric interfaces.
• This method enhances the reliability and consistency of EMG-driven MMs for long-term applications.
• The findings pave the way for more intuitive and dependable human-machine interaction systems.