Analysis of EEG signals in gazing at rotating vanes for brain computer interface

A brain computer interface system (BCIs) is a device that translates brain activity into a command for a computer. This thesis proposes a new BCIs based on the gaze on rotating vanes. Our BCIs can identify five different rotating vanes that were shown to the subjects in a screen. The EEG signals were obtained from healthy human subjects in an age group between 20 and 32 years. The features are extracted from the 0.5-sec, 1-sec. and 2-sec. epochs using different methods. These features by different classifiers were classified and the results were compared together. FFT, DWT and AR model to extract features and SVM, k-NN, LDC and PLSR to classify these features were used. PLSR classifier has better classification acuracy in different steps of thesis. Also channel T3 has better results in gazing rotating vanes. By using only this channel, We could classified 2-sec epochs in proposed spelling system, with about %65. Our system's speed is about 21 bits per min.