A visual evoked potential based brain computer interface

The neurons that form the basis of brain functions, communicate with each other through ionic currents. The electrical activity that occurs as a result of this communication is recorded by a method called electroencephalography. A brain-computer interface design is possible if the electrical activity in the brain is known to be related to cognitive or motor activities. Brain-computer interface is defined as the interface that uses brain signals to control a device or to communicate between the device and the user. Brain-computer interfaces are categorized according to the electrophysiological source used. The electrophysiological source used in this study is the electrical activity recorded from the occipital lobe and called visual evoked potentials. In this study, the usability of computer programs running on standard computer hardware by using visual evoked potentials obtained from the occipital lobe of the brain was investigated. For this purpose, the electroencephalograms obtained from the subjects were analyzed and the classification performances of the brain-computer interface application under different conditions were compared. Electroencephalogram analysis generally consists of digital signal processing, feature extraction and classification stages. In the digital signal processing stage, the effects of the high pass and low pass filters applied to the signals on the classification performance are investigated. As a result of this research, no significant change in the classification performances of designs with only high pass filtering and high and low pass filtering has been observed. In addition, it has been observed that only the designs include a high-pass filtering implementation exhibit better classification performance in many cases. Visual evoked potential based brain-computer interface designs use the spectral properties of the signals for classification. Frequency spectrum of the visual evoked potentials consist of harmonics of integer multiples of the stimulus frequency. In this thesis, the classification performances were compared with the spectral properties extracted from various combinations of the first four harmonics of the stimulus frequencies. As a result, classification performed with features extracted from first, second and fourth harmonics of the visual stimulus frequencies has been found to be more successful than classifications performed with features extracted from other harmonic combinations. In addition, the effects of time window length, overlapping harmonic frequencies of stimuli and different classification algorithms on the performance of brain-computer interface applications have been investigated.