Analysis of Surface EMG Signals to Control of a Bionic Hand Prototype with Its Implementation.

Area of Science:

• Biomedical Engineering
• Robotics
• Machine Learning

Background:

• Surface electromyography (sEMG) signals offer a non-invasive method for detecting muscle activity.
• Accurate interpretation of sEMG is crucial for advanced prosthetic control.
• Existing systems often require complex calibration and feature extraction techniques.

Purpose of the Study:

• To develop a comprehensive system for acquiring and processing multi-channel sEMG data.
• To extract discriminative time-frequency features for hand gesture classification.
• To enable precise control of a simplified bionic hand prosthesis using sEMG signals.

Main Methods:

• sEMG data acquired using an ARM Cortex M3 board and OYMotion sensors.
• Signal processing and feature extraction performed in MATLAB using Fourier and Hilbert-Huang transforms.
• Artificial Neural Networks (ANNs) and machine learning algorithms employed for gesture recognition.
• Experimental protocol involved 109 healthy volunteers performing five distinct hand gestures.
• Bionic hand prototype constructed using 3D printing and actuated by servo motors.

Main Results:

• Identified discriminative time-frequency features from sEMG signals for gesture classification.
• Successfully trained an ANN model for recognizing predefined hand gestures.
• Demonstrated the feasibility of the system for controlling a simplified bionic hand.

Conclusions:

• The developed sEMG acquisition and analysis system effectively extracts features for hand gesture classification.
• The system shows promise for enhancing the control capabilities of bionic hand prostheses.
• Further research can explore more complex gestures and adaptive learning algorithms.