A new method for classification of focal and non-focal EEG signals

Abstract

In this paper, we have proposed a new methodology based on the empirical mode decomposition (EMD) for classification of focal electroencephalogram (FE) and non-focal electroencephalogram (NFE) signals. The proposed methodology uses EMD along with Sharma–Mittal entropy feature computed on Euclidean distance values from K-nearest neighbors (KNN) of FE and NFE signals. The EMD method is used to decompose these electroencephalogram (EEG) signals into amplitude modulation and frequency modulation (AM–FM) components, which are also known as intrinsic mode functions (IMFs) then the KNN approach-based Sharma–Mittal entropy feature has been computed on these IMFs. These extracted features play significant role for the classification of FE and NFE signals with the help of least squares support vector machine (LS-SVM) classifier. The classification step includes radial basis function (RBF) kernel along with tenfold cross-validation process. The proposed methodology has achieved classification accuracy of 83.18% on entire Bern-Barcelona database of FE and NFE signals. The proposed method can be beneficial for the neurosurgeons to identify focal epileptic areas of the patient brain.