Abstract
This main objective of this paper is to develop a fuzzy classification model for epilepsy risk level analysis from EEG signals. The parametric values such as energy, variance and duration, co-variance, peaks, sharp and spike waves and events are derived in each epoch of two second duration in the EEG signal channels. Fuzzy techniques are applied to classify the risk level in each epoch for all the channels, the obtained risk level patterns are found to have low values of sensitivity, specificity Performance Index (PI) and Quality Value (QV). In order to increase the classification rate, a neural network Radial Basis Function (RBF) is used for optimization of fuzzy outputs. This network is trained and tested with 480 patterns extracted from three epochs of sixteen channel EEG signals often known epilepsy patients. Different architectures of RBF networks are compared based on the minimum Mean Square Error (MSE); the better network (16-16-1) is selected. The Quality Value is improved to 23.98 when compared to the value 6.25 achieved through fuzzy techniques.
