Analyzing Haar and dB2 with Compensatory GMM Classifier for Epilepsy Detection

Abstract

Epilepsy is a neurological illness that affects a significant number of individuals all over the globe, and the treatment that they get with medicine is not always effective. Analyzing recordings made using electroencephalography (EEG) could provide one with a wealth of information on the system that is responsible for the formation of epilepsy. For exhibiting the many attributes of non stationary signals, like recurring patterns and discontinuities, the wavelet transform tool is very helpful. Therefore, the wavelet transform tool is employed in order to quantify and investigate the epileptiform events. In this study, Haar and dB2 are employed to reduce the features dimensionality from EEG outputs. After this, the reduced information is identified with the assistance of a Compensatory Gaussian Mixture Model (GMM) learning algorithm. Results indicate that an average accuracy of 89.43% is achieved when the Haar wavelet features is identified using compensatory GMM and an average accuracy of 85.75% is achieved when the dB2 wavelet features is identified using compensatory GMM.