EEG Analysis Using a Wavelet Packet Transforms Mean Energy and Mean Teager Energy with an Artificial Neuro-Fuzzy System

The present study makes use of the energy and entropy measurements of the human brain for detection of brain-related abnormalities. The electroencephalogram (EEG) records the brain’s signals, which contain valuable information about the normal or abnormal state of the brain. Many studies have focused on the nonlinear analysis of EEG mainly for the characterization of abnormal (epileptic) brain states. Here the EEG is first decomposed into four levels using wavelet packet transform. The packets on each level of the decomposition are a linear combination of wavelet basis functions. All the frequency bands, i.e., delta, theta, alpha, beta, and gamma, are determined. Then for the feature extraction vector, the statistical parameters like entropy, energy, mean energy, and mean Teager energy are used. The effects of the frequency rhythm of brain waves are analysed. The abnormal EEG is more affected on the delta and theta frequency bands. The classification is done using the adaptive neuro-fuzzy inference system (ANFIS). It has been observed that mean Teager energy has a minimum training error compared to other parameters.