Artificial intelligence in brain computer interfaces

Includes bibliographical references (page 49)In this investigation, classification of electroencephalographic (EEG) data for Brain Computer Interfaces is studied in order to distinguish the most optimal methods for classifying hand movements. Algorithms previously studied in the field of BCI technology, as well as novel algorithms showing promise were selected for classification and feature extraction. Optimization techniques were used to distinguish which classification algorithms generated the highest prediction accuracy. Feature extraction algorithms were optimized as well in order to determine which algorithm generated the highest reduction in data dimensionality while sustaining or improving performance in classification accuracy. Together, a classification algorithm feature extraction algorithm pair would prove to generate the highest classification accuracy while significantly reducing data dimensionality. This study showed that an optimized Support Vector Machine feature extraction algorithm paired with an optimized Support Vector Machine, using a radial basis function kernel, produced the highest prediction accuracy while generating the highest reduction in dimensionality of the specified EEG data