Presenting a technique for identifying and diagnosing epileptic seizures using nonlinear feature extraction based on DT-CWT coefficients of brain EEG signals with a deep
الموضوعات :
abdullah jafari chashmi
1
(Department of Electrical Engineering, shahmirzad Branch, Islamic Azad University, Semnan, Iran)
الکلمات المفتاحية: epilepsy, Feature reduction, k-Means Algorithm, Nonlinear features, radial basis function networks, brain EEG classification,
ملخص المقالة :
Epilepsy is a type of brain disease that can be diagnosed by observing EEG signals. The disease often occurs in children. However, some cases are also seen in adults. Diagnosing this disease in the early stages is a challenging task for doctors. In this work, the authors have classified epileptic and normal EEG signal by adopting deep learning approach. To achieve the efficient features, the dual tree complex wavelet (DTCWT) is considered. Then, the decomposed wavelet coefficients are applied to nonlinear feature extraction. These features are used as input to the Radial Hybrid Basis Function (RBF) class. Using the proposed method, about 99% classification accuracy is observed. This requires significant improvement of the proposed algorithm compared to other previously presented algorithms. It is the first time that nonlinear feature extraction on DT-CWT coefficients of an EEG signal is used to diagnose epilepsy.
