부산대학교 도서관

Electroencephalography (EEG) plays major role in detecting the activities of the human brain which acts as important method for assisting the epilepsy diagnosis. The electrical activity of the human brain is thought to be a crucial source of information for the detection and classification of epileptic seizures. Here, epilepsy is automatically classified from EEG signals using the multivariate and time series data set characteristics. This dataset consists of raw time domain EEG signals. The EEG signals from these five classes are obtained from the UCI repository and include EEG signals recorded with the eyes open, closed, from healthy brain regions with identified tumour regions, from the tumour location, and from EEG signals from healthy brain regions and EEG recorded with seizure activities. The total of 1750 samples collected from the UCI repository and analyzed with the four groups of classifiers as CNN, GRU, LSTM and CNN-LSTM. The CNN-LSTM achieves the significant accuracy of 98% compared to the other classifiers. It is concluded that CNN-LSTM achieves the better detection and classification of epileptic seizure.