EEG-Based Motor Execution Classification of Upper and Lower Extremities Using Machine Learning

Abstract

This study classifies upper- and lower-extremity motor execution from electroencephalography (EEG). We compared two feature extractors, statistical features and Common Spatial Patterns (CSP), and four classifiers: K-Nearest Neighbors, Linear Discriminant Analysis (LDA), Multilayer Perceptron, and Support Vector Machine. Metrics were accuracy, F1, precision, and recall. CSP with LDA achieved the best, most consistent performance (72.5% accuracy); statistical features underperformed. We report real-time feasibility benchmarks, post-cue time-window analysis, and significance tests for classifiers. Findings support BCI and neuroprosthesis development, while noting subject variability and dataset specificity. Future work is real-time use, cross-dataset generalization, and hybrid deep learning.