Unveiling EMG semantics: a prototype-learning approach to generalizable gesture classification.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Prosthetics

Background:

• Upper limb loss significantly affects quality of life and daily function.
• Electromyography (EMG) signal decoding is crucial for restoring limb function.
• Current EMG-based gesture classification methods lack subject-to-subject generalizability.

Purpose of the Study:

• To present a novel deep prototype learning method for accurate and generalizable EMG-based gesture classification.
• To overcome limitations of existing methods in cross-subject generalization.
• To improve the reliability and precision of EMG gesture recognition for upper limb prosthetics.

Main Methods:

• Implemented a deep prototype learning framework for EMG signal analysis.
• Developed a method that matches new EMG inputs to learned prototypes for label prediction.
• Validated the approach on four Ninapro datasets.

Main Results:

• The deep prototype learning method significantly enhanced classification performance and generalizability.
• The classifier demonstrated superior intra-subject and inter-subject accuracy compared to state-of-the-art methods.
• Subtle differences between gestures were effectively discriminated, increasing reliability.

Conclusions:

• The proposed deep prototype learning method is effective for EMG gesture classification.
• This approach shows promise for advancing upper limb prosthetic control.
• The findings pave the way for more seamless and accurate prosthetic functionality.