Cursor Control System: A Brain-Computer Interface Using EEG Signals and DBN-Based Classification Method

Abstract

Since the invention of computer, it and its related technologies have been prosperously expanded and developed with an incredible pace. Nowadays, computer has penetrated into almost everywhere as indispensable tool. Electroencephalography (EEG) has attracted significantly more and more attention from researchers and scientists, because it has built up a pathway for researchers and scientists to help explore and understand human brainÍs anatomy, working principles, brain related diseases and so forth. It has been verified that the brainÍs commands to body actions are transmitted in form of electricity, which is the source of EEG signals. Therefore, the EEG signals could reflect corresponding physical actions from human body. In the other words, that means recording and understanding EEG signals could possibly foresee human intentions and actions. What will occur if we can merge EEG signals and computer together? Can we generate affective computer? Such questions have gathered a deal of interests in past decades. Brain-Computer Interfaces (BCIs), which stands on the intersection of EEG and computer, has become a popular research field based on same reason. A cursor control system, which is manipulated by EEG signals, has been proposed in this work. It is designed to monitor EEG signals and translate them into the commands of cursor control, including 2-dimentional cursor movement and 4 functional cursor manipulations, selecting, dragging, opening and deleting. In order to create a high efficient and accurate system, Deep Belief Nets (DBN) has been adopted as the kernel of classification methodology.