Classification and discrimination of focal and non-focal EEG signals using hybrid features and support vector machine

Abstract

In the current scenario, computerised epileptic seizure detection is an emerging research area in the field of medical diagnosis. In this paper, electroencephalogram (EEG) signals were collected from the dataset: Bern-Barcelona EEG. After collecting the EEG signal, decomposition was performed using empirical mode decomposition (EMD), which has the ability to determine the subtle changes in frequency. Then, hybrid feature extraction was performed using Renyi entropy and Teager energy operator (TEO) for extracting the features from collected EEG signals. To obtain optimal feature subset, a dimensionality reduction approach; principal component analysis (PCA) was employed. After dimensional reduction, an effective supervised classifier; support vector machine (SVM) was used to classify the seizure and non-seizure EEG signals. The experimental outcome showed that the proposed approach distinguishes the non-seizure and seizure EEG signals using the performance measures like sensitivity, accuracy, specificity, f-score and Matthew's correlation coefficient (MCC). The proposed methodology improves the classification accuracy in epileptic seizure detection up to 7%-30% compared to the existing epilepsy seizure methods.