sEMG feature selection and classification using SVM-RFE.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Human-Computer Interaction

Background:

• Classifying hand movements using surface electromyography (sEMG) data is challenging.
• Prior research focused on signal processing, feature extraction, and classification algorithms to enhance accuracy.
• Effective feature selection for wrist and finger movement classification remains an underexplored area.

Purpose of the Study:

• To introduce and evaluate a feature selection step within the classification pipeline for hand movement classification.
• To enhance classification accuracy and performance by optimizing the feature set used for training.
• To investigate the novelty and impact of feature selection in the specific domain of wrist and finger movement analysis.

Main Methods:

• Implemented a feature selection algorithm to identify the most relevant features from a larger set.
• Utilized surface electromyography (sEMG) data recorded during hand movements.
• Trained and evaluated classification models using the selected subset of features.

Main Results:

• Achieved a significant reduction in the number of features required for classification (from 144 to 39-53).
• Demonstrated substantial performance gains in classification tasks.
• Obtained promising classification accuracy, exceeding 90% for certain subjects.

Conclusions:

• Feature selection is a valuable addition to the sEMG-based hand movement classification process.
• The proposed method significantly enhances classification performance and accuracy.
• This approach offers a novel and effective strategy for improving human-computer interaction systems reliant on hand gesture recognition.