Comparison of Seizure Detection Performances of Features Based on Wavelet Transform and Empirical Mode Decomposition

Several features are used in order to evaluate the epileptic components of the electroencephalogram (eeg) signals. the generated feature matrices are applied to different classifiers as input. it is aimed to detect different epileptic stage. in this study, performances of wavelet transform and empirical mode decomposition methods which are used commonly to extract feature in epilepsy studies have been compared. eeg signals, which contain normal and seizure stages, have been divided into 5 sub-bands including different frequency components via both methods. feature matrices have been obtained by calculating mean, standard deviation, entropy and power for each sub-band. the feature matrices have been classified by k-nearest neighbor algorithm and results have been compared for both feature extraction methods. analysis has been implemented patient-specifically for 14 patients with epilepsy.