Genetic Algorithm Application to Feature Selection in sEMG Movement Recognition with Regularized Extreme Learning...

Area of Science:

• Biomedical Engineering
• Machine Learning
• Rehabilitation Robotics

Background:

• Surface electromyography (sEMG) signals are crucial for understanding and predicting hand-arm movements.
• Effective feature selection is vital for accurate sEMG-based movement classification.
• Existing methods may not optimally select features for individual signal channels.

Purpose of the Study:

• To develop and evaluate a genetic algorithm (GA) based feature selection strategy for sEMG hand-arm movement prediction.
• To identify the optimal feature set for each sEMG channel independently.
• To improve classification accuracy in hand-arm movement prediction tasks.

Main Methods:

• Utilized a genetic algorithm (GA) for feature selection from sEMG data.
• Evaluated features independently for each channel.
• Employed Regularized Extreme Learning Machine (RELM) for the classification stage.
• Tested the approach on the Ninapro database 2, exercise B, considering 11 time-domain and 2 frequency-domain metrics.

Main Results:

• Achieved a classification accuracy of 87.7%.
• Selected an average of 43 combined feature/channel.
• Demonstrated promising results compared to previous studies in sEMG-based movement prediction.

Conclusions:

• The proposed GA feature selection strategy is effective for sEMG hand-arm movement prediction.
• Independent channel feature evaluation enhances classification performance.
• The approach offers a robust and accurate method for prosthetic control and human-computer interaction.