ABSTRACT
Epilepsy affects about fifty million people worldwide and is characterized by abnormal and sudden hyper-synchronous excitation of the neurons in the brain. The electroencephalogram (EEG) is the most widely used method to record brain signals and is widely used in epilepsy diagnosis. Traditional signal decomposition methods, such as Empirical Mode Decomposition (EMD) and Hilbert-Huang transform used for classification are recursive in nature, making them more susceptible to noise and chosen sampling rates. In this chapter, we use the method of Variational Mode Decomposition (VMD) to classify seizure/seizure free signals. This technique uses variational non recursive mode decomposition, which decomposes a signal into components called principal modes. Four main features are extracted from these principal mode components, namely Renyi Entropy, second order difference plot (SODP), fourth order difference plot (FODP) and average amplitude which are then used for the classification, both individually and considering all four features. The classification was done using a Multilayer Perceptron (MLP) architecture with back propagation algorithm used for training. An average accuracy of 97.5% was obtained when the features were used individually and an accuracy of 99.1% was obtained when all the features were considered, which is higher than the accuracies obtained by Empirical Mode Decomposition (EMD) and Hilbert-Huang transform methods for the same dataset. The proposed approach was tested on the Bonn University EEG signal dataset.
