ADADELTABASED LEARNING RATE SCHEDULER WITH DEEP NEURAL NETWORKS FOR EPILEPTIC SEIZURE DETECTION AND CLASSIFICATION MODEL

Electroencephalogram (EEG) signal based epileptic seizure identification and classification becomes an important research domain, which determines the nonstaticgrowth of brain actions. Generally, epilepsy detection can be done by training professionals by examining the EEG signals, which is time consuming and noise sensitive. In recent times, the advent of deep learning (DL) models can be employed to detect and classify the presence of epileptic seizures from EEG signals. This paper develops new deep learning with learning rate scheduler based epileptic seizure detection and classification model. The DNN-AD method performs preprocessing, classification and learning rate scheduling processes. For classification purposes, a deep neural network (DNN) model is employed, which identifies the presence of multiple class labels of seizure. Besides, the DNN-AD model also includes the Adadelta technique to determine the learning rate of the DNN model, which results in improved classification performance and reduced training time. In order to examine the effective performance of the DNN-AD model, the Epileptic Seizure Recognition dataset is used. The simulation results demonstrate the superior performance of the DNN-AD model over the compared methods with the maximum accuracy of 89.08% and 92.9% under the classification of binary and multiple classes respectively.