Self-organized clustering approach for motion discrimination using EMG signal

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Signal Processing

Background:

• Controlling myoelectric hands requires accurate discrimination of electromyography (EMG) signals.
• Overlapping EMG feature patterns from different motions pose a significant challenge for current classifiers.
• Existing methods struggle to achieve high discrimination rates, limiting prosthetic hand functionality.

Purpose of the Study:

• To develop an improved motion discrimination method for myoelectric hand control using EMG signals.
• To address the issue of overlapping EMG feature patterns that hinder accurate classification.
• To enhance the precision and reliability of prosthetic hand movement intention detection.

Main Methods:

• Utilized a self-organized clustering method to extract representative EMG feature patterns.
• Assigned class labels to feature patterns based on hand and finger joint angles for intended motions.
• Trained a classifier using extracted feature patterns and corresponding class labels.

Main Results:

• The proposed method demonstrated a 5-30% higher discrimination rate compared to conventional methods.
• Experimental results validated the effectiveness of the novel motion discrimination technique.
• Improved accuracy in classifying intended motions from EMG signals was achieved.

Conclusions:

• The developed method significantly enhances the discrimination of EMG signals for myoelectric hand control.
• This advancement offers a more effective solution for prosthetic hand operation by improving motion classification.
• The findings support the potential for more intuitive and precise control of myoelectric prosthetics.