Authors: S. Aarthishree; M. Jayashree; J. Rhinose Fathima
Addresses: Department of Information and Communication Engineering, Anna University Regional Centre, Coimbatore, India ' Department of Information and Communication Engineering, Anna University Regional Centre, Coimbatore, India ' Department of Electronics and Communication Engineering, Anna University Regional Centre, Coimbatore, India
Abstract: Brain, one of the essential organs in humans is likely to be affected by degenerative disorder, known as epilepsy. Epileptic seizure is caused by deviations in electrical activity of certain brain cells. Usual diagnostic tests are EEGs and brain scans, which are cost effective. Here, the electroencephalogram signal plays a major part in diagnosing epilepsy. But the detection needs the analysis of whole EEG with respect to time. Here, the ultimate aim is to make the process more accurate and fast in detection by limiting the number of data points through setting threshold limits. It allows reduction in data points to obtain denoised signal. Process flow: EEG signal is denoised by performing the integration process of wavelet transform and Otsu threshold process. By applying the inverse wavelet transform original signal get obtained. Then by sample entropy their feature gets extracted and is used along with extreme learning machine model.
Keywords: electroencephalograms; EEG signals; sample entropy; wavelet transform; extreme learning machine; ELM; epilepsy detection; brain seizures; epileptic seizures; signal denoising; feature extraction.
International Journal of Advanced Intelligence Paradigms, 2016 Vol.8 No.4, pp.412 - 424
Received: 12 Nov 2014
Accepted: 17 May 2015
Published online: 07 Nov 2016 *