Stable force-myographic control of a prosthetic hand using incremental learning.

Area of Science:

• Biomedical Engineering
• Rehabilitation Engineering
• Machine Learning in Prosthetics

Background:

• Force myography (FMG) is a potential alternative to electromyography (EMG) for upper limb prosthesis control.
• A key challenge in FMG is the non-stationary nature of force data, which varies with prosthesis orientation and position.
• This variability can degrade the performance of control systems over time.

Purpose of the Study:

• To propose and evaluate an incremental learning method to address the non-stationarity of force myography signals.
• To improve the robustness and adaptability of FMG-based control for upper limb prostheses.
• To demonstrate the effectiveness of continuous adaptation for maintaining performance in changing conditions.

Main Methods:

• Utilized an online sequential extreme learning machine (OS-ELM) for incremental learning.
• Implemented occasional updates to allow the model to continually adapt to signal variations.
• Tested the method for predicting hand status from forearm muscle forces across various arm positions.

Main Results:

• The incremental learning approach effectively adapted to changes in force myography signals.
• Consistent performance was maintained throughout the testing period.
• Achieved an average classification accuracy of 98.75% for two subjects in predicting hand status.

Conclusions:

• Incremental updates are effective in maintaining stable performance for force myography-based prosthetic control.
• The proposed online sequential extreme learning machine method offers a robust solution to signal non-stationarity.
• This approach enhances the reliability and usability of upper limb prostheses utilizing force myography.