A new method for diagnosing epilepsy using dictionary learning

Abstract

Epilepsy is a disorder of the central nervous system. An electroencephalograph is often used to diagnose epilepsy. In this study, we aim to diagnose epilepsy from the EEG signals using a new method of dictionary learning and sparse coding. In the pre-processing, Butterworth and notch filters are used to remove noises, K-singular value decomposition (K-SVD) algorithm is used to learn a dictionary to find a matrix of dictionary atoms, and in sparse coding, the orthogonal matching pursuit (OMP) algorithm is used to extract the features from the signals. The extracted features were entered as input for classification of signals into two groups of epileptic and non-epileptic signals, using the support vector machine (SVM) method. The results obtained in this method have an accuracy of 97.89%, higher than other methods, due to its excellent training by K-SVD and feature extraction, which is well done by OMP.