Classification of Executed and Imagined Motor Movement EEG Signals

Electroencephalography (EEG), which contains cortical potentials during various mental processes, can be used to provide neural input signals to activate a brain machine interface (BMI). The effectiveness of such an EEG-based prosthetic system would rely on correct classification of executed motor signals from imagined motor movement signals; an executed motor signal should initiate movement in the artificial limb while signals from motor imagery should be filtered out. This work evaluates the performance of features based on average EEG signal power contained in different frequency bands in order to distinguish between the executed and imagined EEG signals. We also investigate Independent Component Analysis (ICA) as a scheme to remove irrelevant artifacts from the EEG signals. Results demonstrate that using EEG for classification can be performed effectively; however results vary significantly from patient to patient, suggesting that BMI must highly specialized for an individual patient.