EEG-based motor execution classification of upper and lower extremities using machine learning

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Electroencephalography (EEG) is crucial for brain-computer interfaces (BCIs).
• Accurate classification of motor execution is vital for neuroprosthetics and assistive technologies.
• Evaluating different feature extraction and classification methods is essential for BCI performance.

Purpose of the Study:

• To classify upper- and lower-extremity motor execution using EEG signals.
• To compare the efficacy of statistical features versus Common Spatial Patterns (CSP) for feature extraction.
• To evaluate four distinct machine learning classifiers: K-Nearest Neighbors, Linear Discriminant Analysis (LDA), Multilayer Perceptron, and Support Vector Machine.

Main Methods:

• Feature extraction using statistical methods and CSP.
• Classification employing K-Nearest Neighbors, LDA, Multilayer Perceptron, and SVM.
• Performance evaluation using accuracy, F1 score, precision, and recall metrics.

Main Results:

• Common Spatial Patterns (CSP) combined with Linear Discriminant Analysis (LDA) demonstrated superior and consistent performance, achieving 72.5% accuracy.
• Statistical features extraction methods underperformed compared to CSP.
• Real-time feasibility benchmarks and significance tests were conducted.

Conclusions:

• The combination of CSP and LDA is highly effective for classifying motor execution from EEG data.
• Findings support the advancement of BCI and neuroprosthesis development.
• Subject variability and dataset specificity are important considerations for future BCI applications.