The aim of the Electroencephalography signal analysis to lead the creation of useful systems. This paper designs an experiment to classify the motor imagery EEG data using four different features and support vector machines. The motor imagery EEG data obtained from the publicly available BCI competition III. The data set consists of cued motor imagery (multi-class) with four classes (left hand, right hand, foot, tongue) three subjects. The recording was made with a 64-channel EEG amplifier from Neuroscan, using the left mastoid for reference and the right mastoid as ground. The study focuses on the classification of left-hand and right-hand movements. So, the data about left-hand and right-hand movements are extracted from the competition data set. Then, the feature extraction phase is performed to trails of the data set. The extracted features are mean absolute value (MAV), waveform length (WL), Variance (VA) and slope sign changes (SSC). The feature matrix is composed of combining the four features obtained from every trail of the EEG data. For the dimensionality reduction and eliminating the redundant data, Principal component analyze (PCA) is employed for the feature selection. The final data set is served to the support vector machine (SVM) classifier to discriminate motor imagery EEG data.
Anahtar Kelimeler: EEG, Motor Imagery BCI Data, Mean Absolute Value (MAV), Waveform Length (WL), Variance (VA) and Slope Sign Changes (SSC), SVM