Automatic Seizure Detection of EEG Signals Using Deep Belief Network With Robust LMD-Based Features

Automatically identifying epileptic activity using electroencephalogram (EEG) data is critical in treating seizures. To this end, the study employs a novel analytical time-frequency algorithm called robust local mean decomposition (RLMD). An arbitrary signal can be divided into multiple product functions (PFs) by RLMD. This study proposes a unique method to predict seizures that combines RLMD with deep convolutional neural networks (DCNNs). First, The EEG signals are decomposed into a series of product functions by RLMD. Next, three PFs (PF2–PF4) are selected and applied to DCNN to automatically learn the EEG features. Using the publicly accessible EEG dataset from Bonn University, the proposed approach achieves sensitivity, specificity, and accuracy of 98.1%, 98.4%, and 98.5%, respectively. According to the simulation findings, the proposed strategy may be used in clinical settings to predict epileptic seizures.