Feedback-aided data acquisition improves myoelectric control of a prosthetic hand

Area of Science:

• Biomedical Engineering
• Machine Learning
• Rehabilitation Technology

Background:

• Pattern-recognition-based myocontrol for upper-limb prostheses often suffers from poor generalization.
• Unreliable myocontrol limits clinical adoption and everyday use of prosthetic devices.
• Machine learning model performance is hindered by poor generalization to untrained conditions.

Purpose of the Study:

• To investigate the impact of interactive, feedback-aided data acquisition on myocontrol performance.
• To compare novel feedback-guided protocols against standard data acquisition methods.
• To enhance the reliability and usability of upper-limb electric prostheses through improved training.

Main Methods:

• 18 non-disabled participants evaluated three data acquisition protocols: two feedback-aided and one standard.
• Protocols dynamically acquired data across multiple arm positions to address limb position effects.
• Feedback-aided protocols used acoustic signals to guide participants in acquiring data in critical regions.

Main Results:

• Feedback-aided data acquisition significantly outperformed the unguided standard protocol.
• Objective and subjective measures indicated superior performance with feedback-guided methods.
• The study demonstrated improved myocontrol through user interaction during data acquisition.

Conclusions:

• Interactive data acquisition is fundamental for enhancing myocontrol.
• Feedback-guided training protocols improve the generalization and reliability of myoelectric control systems.
• User involvement in the training process is key to advancing prosthetic hand functionality.