sEMG-Based Gesture Classifier for a Rehabilitation Glove.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Machine Learning

Background:

• Surface electromyography (sEMG) signals are crucial for controlling prosthetic and rehabilitation devices.
• Accurate hand gesture recognition directly impacts the effectiveness of these control mechanisms.
• Existing methods require improvement for real-time, user-specific applications.

Purpose of the Study:

• To develop a highly accurate classifier for hand gesture recognition from sEMG signals.
• To create an efficient interface for real-time control of rehabilitation devices.
• To enable user-specific adaptation of the gesture recognition algorithm.

Main Methods:

• A novel classifier combining Bayesian neural networks, pattern recognition networks, and layer recurrent networks was developed.
• An efficient interface was designed for real-time classification and user retraining.
• The system integrates with shape memory alloy-based rehabilitation devices.

Main Results:

• The proposed classifier achieved 98.7% accuracy in online hand gesture recognition from sEMG data.
• The developed interface allows users to retrain the algorithm with their own data in minutes.
• The system successfully detects user movement intention for real-time device control.

Conclusions:

• The novel classifier offers a promising solution for accurate sEMG-based hand gesture recognition.
• The user-trainable interface enhances the adaptability and usability of rehabilitation devices.
• This approach effectively bridges the gap between user intention and device control.