Classification of Seizure Prone EEG Signal Using Amplitude and Frequency Based Parameters of Intrinsic Mode Functions
Epilepsy is a serious neurological disorder which disturbs the common activities of a human brain. It is detected with the help of electroencephalogram (EEG) recording of the brain as it covers detailed information related to the different physiological states of the brain. A novel method based on empirical mode decomposition (EMD) with amplitude and frequency based parameters is presented in this paper to classify the seizure activity from an EEG signal. A well-known dataset consisting of EEG signals of healthy volunteers and epileptic patients, under different stages, has been used in this study. EMD has been employed for the decomposition of an EEG signal into a set of intrinsic mode functions (IMFs). First four IMFs have been considered in this work. Two amplitude based parameters and one frequency based parameter have been extracted from each of the four IMFs. The validation of the extracted parameters of all IMFs has been carried out by Kruskal–Wallis statistical test and box plot method. Extracted parameters qualify both the validation tests for discriminating normal, interictal and ictal EEG signals. Twelve features from each EEG signal have been fed to four different classifiers; artificial neural network (ANN), least square-support vector machine (LS-SVM), random forest and Naïve Bayes. The performance of the study has been evaluated in two stages. In the first stage, parameters from first four IMFs have been analyzed individually, with four classifiers and in the second stage, the performance has been analyzed by considering the parameters of all the four IMFs. The results of the proposed method have been compared with the other state of the art techniques for the validation.