Analysis of EEG signals by eigenvector methods

Electroencephalography (EEG) is the most important tool to study the brain behavior. This paper presents an integrated system for detecting brain changes during relaxation and writing condition. In most studies, which use quantitative EEG analysis, the properties of measured EEG are computed by applying power spectral density (PSD) estimation for selected representative EEG samples. The power density spectra were calculated using three different eigenvector methods namely Pisarenko, Multiple Signal Classification (MUSIC) and Modified Covariance. The statistic features were calculated for each segmented signal and Principle Component Analysis (PCA) was implemented in order to reduce the feature vector dimension. The PSD values obtained by PCA were used as inputs of kNN classifier. The classification results showed that Modified Covariance is the most suitable features to discriminate relaxation and writing task with the average accuracy of 95%. It confirmed that the features have potential in detecting the electroencephalographic changes.