Room No.- ECE 1110, Electronics and Communication Engineering Discipline, Dr. Syttendranath Academic Building (Academic Building 1), Khulna University, Khulna 9208.




    Personal Webpage:
    click here

Feature Extraction and Classification of EEG signals for Seizure Detection

Abstract: Epilepsy Is a neurological disorder marked by sudden recurrent episodes of sensory disturbance, loss of consciousness or convulsions which is associated with electrical activities pf the brain It is a brain disorder which affects people of all ages. It is characterized by unpredictable seizures and can cause any other health problem. It is one of the most common neurological disorder which affects 1% of the whole world population and among them about 0.2% affected individuals lose their lives. A report reveals that approximately 50 million of worldwide have epilepsy and approximately 138 million people are suffering from epilepsy in our country. Mostly 1 in every 200 newborn babies face some form of seizures. The psychological consequence of epilepsy and the permanent state of being epileptic are very problematic. It effects on social, cultural and personal factors The social consequence of epilepsy is very severe and included with shorten lifespan, excessive bodily injury, social disability and quality of life is hampered in epilepsy. This research presents an autonomous system, capable of detecting the occurrence of an epileptic seizure, without the help of an expert The proposed system consists of four steps i.e. pre-processing, feature extraction, feature selection and classification. The purpose of pre-processing is to organize the data in an orderly manner and to remove noise. The feature extraction step is required in the proposed. Different time, frequency, time frequency (wavelet) domain features such as power, skewness, kurtosis etc. in four different EEG bands (delta, theta, alpha, beta) have been extracted. In addition, entropy based, and connectivity based have also been extracted. The system then performed the process of feature selection, where the best set of features are determined using Multi-Objective Evolutionary Algorithm and finally used for classification of EEG signals using optimized Support Vector Machine (SVM). Linear The Discriminant Analysis (LDA) and Quadratic Linear Discriminant Analysis (QLLDA) The Analysis (QLDA) hyperparameter of SVM has been performed using Bayesian optimization algorithm The proposed system is tested on a publicly available CHB-MIT dataset and results show the s1gnificance of the proposed system. The optimized SVM with Radial Basis function (RBF) kernel with selected features has achieved an accuracy of 96.96% whereas LDA and QLDA achieved an accuracy of 75.57%, 82.73%, respectively

Role Co-Supervisor
Class / Degree Bachelor

Apu Nandy [140919] , S.M. Nasim Uddin [140928], and  Shafiqul Alam [140933]

Start Date 6th June, 2017
End Date 10th July, 2018