A new framework for classification of multi-category hand grasps using EMG signals

Area of Science:

• Biomedical Engineering
• Signal Processing
• Rehabilitation Technology

Background:

• Electromyogram (EMG) signals are crucial for prosthetic devices, human-machine interfaces, and clinical applications.
• Accurate hand-grasp classification using EMG is essential for improving prosthetic hand functionality and user control.
• Existing methods require enhancement for precision and efficiency in real-world applications.

Purpose of the Study:

• To design an EMG signal-based expert model for accurate hand-grasp classification.
• To enhance prosthetic hand movements for individuals with disabilities through improved control.
• To introduce an innovative framework for recognizing hand movements using EMG signals.

Main Methods:

• A hybrid framework combining logarithmic spectrogram-based graph signal (LSGS) analysis, ensemble feature selection (FS), and an AdaBoost k-means (AB-k-means) classifier.
• LSGS model for extracting relevant features from EMG signals.
• Ensemble FS to identify the most influential features, followed by AB-k-means for classification of hand grasps.

Main Results:

• The proposed LSGS-AB-k-means model achieved a high classification rate on a public EMG hand movement dataset.
• The model demonstrated superior performance compared to several state-of-the-art algorithms.
• The framework proved effective in accurately classifying EMG hand grasps.

Conclusions:

• The developed EMG signal-based expert model accurately classifies hand grasps.
• This model can be implemented as a low-cost, high-classification-rate control unit for prosthetic devices.
• The study advances EMG-based control for rehabilitation and human-machine interaction.