Eeg signal analysis for brain computer interface applications

This thesis investigates the use of EEG signal for controlling a neuro-prosthetic device. The primary aim of this study was to investigate feature extraction and selection methods for classifying this EEG data into two classes namely moving hands and relaxed state. In the first part of the study, the features of EEG signals were extracted using band power in delta (1-3 Hz), theta (4-7 Hz), alpha (8-12 Hz), beta (13-30 Hz) ve gamma(31-50 Hz) bands. Then the feature vector was classified using and without using feature selection methods. In the second part of the study the feature extraction part was repeated using a narrower band range (0-4 Hz, 4-8 Hz, …, 44-48 Hz). Again the feature vector was classified using and without using feature selection methods. The aim was to investigate the effect of using different range of features and feature selection methods to classification performance. The results have shown that using a narrower range for extracting band power and feature selection methods have improved the classification performance of the EEG data.