A robust myoelectric pattern recognition using online sequential extreme learning machine for finger movement...

Area of Science:

• Biomedical Engineering
• Rehabilitation Engineering
• Signal Processing

Background:

• Robust myoelectric pattern recognition is crucial for real-world applications.
• Electrode shift significantly degrades system performance in daily use.
• Existing systems often fail to adapt to these real-world changes.

Purpose of the Study:

• To develop an adaptive myoelectric pattern recognition system.
• To address the challenge of electrode shift and other performance-degrading factors.
• To maintain high system accuracy in practical, everyday scenarios.

Main Methods:

• Implemented an adaptive system using online sequential extreme learning (OS-ELM).
• Trained the OS-ELM model incrementally on data chunks.
• Simulated real-world conditions including electrode displacement.

Main Results:

• The adaptive system maintained an average accuracy of approximately 85%.
• The non-adaptive system showed a significant performance decrease under similar conditions.
• OS-ELM enabled efficient learning from limited data.

Conclusions:

• The proposed adaptive system effectively overcomes electrode shift issues.
• Online sequential extreme learning provides a robust solution for myoelectric systems.
• The adaptive approach ensures sustained performance in practical applications.