Performance Analysis of Morphological Operators Based Feature Extraction and SVD, Neural Networks as Post Classifier for the Classification of Epilepsy Risk Levels

Abstract

Most research to date using hybrid Models focused on the Multi-Layer Perceptron (MLP). Alternative neural network approaches such as the Radial Basis Function (RBF) and Elman network, and their representations appear to have received relatively little attention. Here we focus on Singular value Decomposition (SVD), RBF and Elman network as an optimizer for classification of epilepsy risk levels obtained from code converter using the EEG signals parameters which are extracted by morphological operators. The obtained risk level patterns from code converters are found to have low values of Performance Index (PI) and Quality Value (QV). These neural networks are trained and tested with 960 patterns extracted from three epochs of sixteen channel EEG signals of twenty known epilepsy patients. Different architectures of MLP, Elman and RBF networks are compared based on the minimum Mean Square Error (MSE), the better networks in MLP (16-16-1) and RBF (1-16-1) are selected. RBF out performs the MLP, Elman network, SVD Technique and code converter with the high Quality Value of 25 when compared to the Quality Values of 23.03, 22.2, 20.62 and 12.74 respectively.