Automatic Focal Eplileptic Seizure Detection in EEG Signals

Abstract

In this paper, we propose an automatic epilepsy diagnosis based on the statistical feature extraction. At outset, EEG signals are recorded from the patient and pre-processed to remove the unwanted signals: Dc drift elimination, high pass and low pass filter techniques are applied to preprocess the EEG signals. The noise is diminished from the signal by the method of Hilbert-Huang Transform (HHT). Empirical mode decomposition is the portion of HHT by which intrinsic mode functions (IMFs) are separated from the signal. In Hilbert spectral analysis, the instant frequency of IMFs is executed using Hilbert transform, which allows the finding of localized features. Empirical wavelet transform (EWT) is applied to acquire EWT components from the EEG signals. These features are further extracted in to five frequency subbands based on clinical interest. Genetic algorithm is structured for displaying the best features from the localized features. Based on the optimized features, support vector machine is applied to classify and evaluated the signals as epileptic seizure and seizure-free EEG signals. An experimental result shows that the proposed method can attain a very high accuracy.