Analysis of brain - computer interaction data and related applications

The brain which is the central control mechanism of body, has a complex structure. Many researches have been made to make dynamic structure of human brain more understandable with many different methods Electroencephalography (EEG), is one of these methods able to visualize cerebral activity. Brain – Computer Interface is a developing field about which there are not many researches. The idea of being able to communicate with technological devices via thoughts causes new scientific horizons. The device named Emotiv Epoc+, used in this study, is one of the tools benefited for making EEG signals readable. Cerebral activities are observed via this device thanks to 14 distinct electrodes placed upon proper specialized locations according to different lobes on brain. EEG data used in this thesis study, has been acquired by utilizing mentioned tool and software. Moreover, test of classification algorithms used with purpose of supporting study has been applied to verified EEG data obtained by UCI data warehouse. EEG signals are emerged of collusion of the waves generated in brain. EEG data transmitted to a computer via utilized device and software are written to a file. In classification phase, K-Nearest Neighborhood, C x k- Nearest Neighborhood and Naive Bayesian are utilized. As distance measuring methods of these classification algorithms, Euclidean, Bray-Curtis, Hellinger and Cosine Similarity Metrics take place. In this study, instantaneous EEG signals supplied by the same subject are used. A computer interface in which four main directions are shown is presented to the subject during an experiment. Imagination of movement of the mouse cursor to only one of these four directions is asked from subject. This thinking process of subject is recorded with the help of an EEG measurement device. Each thought of a direction is labeled with respective direction name and all signals are divided into four different classes vii with respect to their labeled direction names. EEG signals grouped in this manner are classified using distinct classification algorithms and classification accuracy rates of those algorithms are compared. Quite more successful algorithms are analyzed by being applied to EEG data in which unknown directions are thought among these four directions.