Seizure detection in neonatal EEG signals using EMD based features

In this paper, an algorithm for epileptic seizure detection using empirical mode decomposition (EMD) method is proposed. The EMD technique decomposes EEG signals into intrinsic mode functions (IMF). Further, statistical features representing seizure and non-seizure EEG activities were computed over these IMFs. The useful features were selected using t-test score and fed to artificial neural network (ANN) classifier to recognizing seizure and non-seizure EEG. The scalp EEG signals recorded from neonatal subjects are used to test the efficacy of the proposed algorithm.