Hand Gesture Recognition Based on High-Density Myoelectricity in Forearm Flexors in Humans

Area of Science:

• Biomedical Engineering
• Machine Learning
• Signal Processing

Background:

• Electromyography (EMG)-based gesture recognition is crucial for decoding fine hand movements.
• Increasing model complexity for accuracy demands extensive data, raising user burden and computational costs.
• Variability in surface EMG (sEMG) signals across users challenges conventional single-feature machine learning approaches.

Purpose of the Study:

• To develop a computationally efficient and accurate method for user-specific gesture recognition using sEMG signals.
• To address the limitations of complex models and single-feature reliance in current gesture recognition systems.
• To improve cross-user pattern recognition performance in EMG-based systems.

Main Methods:

• A novel feature selection method combining mutual information, principal component analysis, and Pearson correlation coefficient (MPP) was proposed.
• The MPP method was designed to identify optimal feature subsets tailored to individual users.
• The selected features were integrated with a Support Vector Machine (SVM) classifier for gesture recognition.

Main Results:

• The proposed MPP feature selection method demonstrated effectiveness in filtering optimal features for specific users.
• Gesture recognition accuracy improved by approximately 5% compared to using single features when employing the optimally selected feature subset.
• The combined MPP and SVM approach achieved accurate and efficient recognition of user-specific gesture movements.

Conclusions:

• The MPP feature selection method offers an effective solution for reducing computational costs in EMG-based gesture recognition.
• This approach enhances the precision of fine hand movement decoding by accommodating user-specific sEMG signal variability.
• The study provides a robust framework for user-specific gesture recognition, paving the way for improved human-computer interaction.