Ensemble Classifier Framework for Epileptic Seizure Classification of EEG Signals

Abstract

A procedure of investigation of Electroencephalogram sign utilizing wavelet change and characterization utilizing AI strategies is created in this research work. EEG sign are non-stationary that makes the visual investigation tedious and may need quantitative examination to uncover shrouded qualities of the signals. Artificial Neural Networks alongside wavelets give capacities to synthesize and analyze the signals and information that shows standard conducts punctuated with unexpected changes. This research work focuses on segregation between two classes of EEG signals; one obtained from healthy persons and other from epileptic patients. This article proposes a technique for analyzing the brain signals, solid extraction of qualities utilizing diverse mother wavelets as Haa r, Coifle t, Daubechies and Sym let. It encompasses arrangement of epilepsy issue utilizing neural systems. A thorough examination focuses to most appropriate mother wavelet to extricate the qualities that further go about as information to the machine learning algorithm. The experimental outcomes got in this exploration work demonstrate that the proposed NN-D ensemble classification strategy yielded greatest grouping precision of 99.4% when contrasted with different other ensembles. The high classification accuracy of the ensemble framework gives clear indication that statistical features obtained from DWT coefficients of the EEG signal yielded a more efficient and reliable solution for differentiation between the epileptic and non-epileptic classes and has a future prospect for classification of other non-stationary biomedical signals. Subsequently, it demonstrates the viability of the proposed strategy for classification of epileptic EEG signals.