Clustering Variational Mode Decomposition for Identification of Focal EEG Signals

The electroencephalogram (EEG) provides a noninvasive way to assess the partial epileptic brain pathology. The diagnosis of epilepsy is made possible by identification of focal EEG signals. In this article, a novel method clustering variational mode decomposition (VMD) is introduced for focal EEG signal identification. VMD is an adaptive signal decomposition technique, which shows the limitations for processing of nonhomogeneous signals, such as EEG. The clustering-VMD (CVMD) is proposed to overcome the limitations of VMD. In CVMD, first, the nonhomogeneous EEG signal is analyzed by optimum allocation sampling into homogeneous EEG-clusters and then decomposed into band-limited modes. The spectral moment based features are extracted from the modes of CVMD and applied as the input to the extreme learning machine classifier for identification of focal EEG signals. The proposed method provides the identification of focal EEG signals, and the performance measures of accuracy, sensitivity, and specificity are 96%, 94.69%, and 97.39%, respectively.