•  
  •  
 

Applied Mathematics & Information Sciences

Author Country (or Countries)

Egypt

Abstract

The detection of epileptic seizures becomes increasingly important because of the widespread of this disease all over the world. Early detection of epileptic seizures helps the patient to manage epilepsy. This paper introduces a detection system for epileptic seizures that implements a Short-Time Fourier Transform (STFT) for denoising the Electroencephalogram (EEG) signal and Wavelet Transform (WT) for features extraction. Four EEG types (i.e Healthy people, Epileptic people during the seizure-free interval (Interictal), Epileptic people during seizure interval (Focal) and Epileptic people during seizure interval (Nonfocal)) are classified by using a Multi-Layer Perceptron Neural Network (MLPNN) classifier. The used dataset was enlarged to be 1050 instead of 500 in the available detection systems. The integration of STFT and WT has an important impact on improving the detection accuracy. The accuracy of the proposed system is 94.4 % which significantly outperforms the previous systems in terms of dataset size and the number of classified EEG signal types.

Digital Object Identifier (DOI)

http://dx.doi.org/10.18576/amis/110616

Share

COinS