Wavelet Neural Networks, Elman Backpropagation and Multilayer Perceptrons for Epilepsy Classification from EEG Signals

The aim of this study is to provide the performance analysis of wavelet networks for the classification of epilepsy from Electroencephalography (EEG) signals. The frequency behaviors of different waveform components of EEG waveforms such as Alpha, Beta, Theta, Delta and Gamma are undertaken. The features are extracted at Wavelet Node and it is given for the neural network classifier inputs. Similarly the code converter features are also extracted and it is given for the neural network classifier inputs. The Multilayer Perceptron Network (MLP) and the Elman Back Propagation Network are the two used networks to identify the patient state for the classification of epilepsy risk levels. The Neural network is trained by Levenberg-Marquardt (LM) algorithm. The Performance Index and the Quality Values are the two important benchmark parameters used for the performance analysis of wavelet networks for classification of epilepsy risk levels from EEG signals.