Resting State EEG Classification for Motor Learning Skills Using Echo State Networks

EEG records the electrical activities from the scalp surface via electrodes. As a modern medical imaging technique, it has been proven to be useful in many different fields. Clinical diagnosis, psychotherapy, brain-computer interfaces and the pharmaceutical industry all have benefited from the insights that one can glean from EEG measurements. However, there exist various difficulties such as uniqueness of individuals, large volume of data and influences of artifacts that prevent us from extracting useful information from those measurements, and thus more involved analytical tools are needed. Recurrent Neural Networks are particularly suitable for dealing with EEG because these networks can capture the critical spatiotemporal characteristics that EEG contains. In this project, we successfully applied Echo State Networks to classify the people’s motor learning skills, given the resting state EEG recording. We also discovered some evidence for the existence of different neurological groups with respect to people’s motor learning skills.