Recognition of Unseen Combined Motions via Convex Combination-based EMG Pattern Synthesis for Myoelectric Control

Area of Science:

• Biomedical Engineering
• Signal Processing
• Rehabilitation Technology

Background:

• Electromyogram (EMG) signals are crucial for intuitive control of assistive devices like prosthetic limbs.
• Collecting extensive training data for all possible motions, especially complex combined ones, is a significant challenge in EMG-based motion recognition.

Purpose of the Study:

• To propose an efficient method for recognizing combined motions using synthetic electromyogram (EMG) data.
• To reduce the burden of data collection by generating combined motion data from basic motion patterns.

Main Methods:

• Developed a technique to generate synthetic EMG data via convex combinations of measured basic motion patterns.
• Utilized both measured basic motion data and synthetic combined motion data for training motion recognition models.
• Conducted an upper limb motion classification experiment with eight subjects to validate the method.

Main Results:

• The proposed method significantly improved the classification accuracy for unseen combined motions.
• An approximate 17% increase in classification accuracy for combined motions was observed compared to traditional methods.
• Demonstrated the effectiveness of synthetic EMG data in expanding the range of recognizable motions.

Conclusions:

• Synthetic EMG data generation through convex combinations offers an efficient solution for training motion recognition systems.
• This approach minimizes the need for extensive real-world data collection, making EMG-based control more accessible.
• The findings pave the way for more sophisticated and intuitive control of assistive devices using EMG signals.