Feature Extraction of Surface Electromyography Using Wavelet Weighted Permutation Entropy for Hand Movement...

Area of Science:

• Biomedical Engineering
• Signal Processing
• Rehabilitation Technology

Background:

• Surface electromyography (sEMG) signal feature extraction is crucial for advanced myoelectric prosthesis control.
• Current methods face limitations in accurately interpreting complex sEMG signals for intuitive prosthetic hand function.
• Improving the practicability of myoelectric prosthetic hands requires more robust and accurate feature extraction techniques.

Purpose of the Study:

• To propose and evaluate a novel feature extraction method, wavelet weighted permutation entropy (WWPE), for sEMG signals.
• To enhance the performance and accuracy of myoelectric prosthetic hand control.
• To compare the effectiveness of the WWPE method against traditional feature extraction techniques.

Main Methods:

• Utilized wavelet transform to decompose and preprocess sEMG signals into distinct wavelet sub-bands.
• Extracted weighted permutation entropies (WPEs) from these sub-bands to form the WWPE feature set.
• Employed support vector machine (SVM) and backpropagation neural network (BPNN) classifiers to recognize seven distinct hand movements using the WWPE features.

Main Results:

• The proposed WWPE feature extraction method demonstrated superior recognition accuracy compared to single sub-band and time-domain feature sets.
• Achieved a maximum recognition accuracy rate of 100% for hand movements.
• Attained average recognition accuracy rates of 100% for SVM and 98% for BPNN classifiers.

Conclusions:

• The WWPE method offers a significant advancement in sEMG feature extraction for myoelectric prosthetics.
• This technique leads to highly accurate recognition of hand movements, improving prosthetic control.
• The findings suggest WWPE is a promising approach for developing more practical and effective myoelectric prosthetic hands.