Improved Chicken Swarm Optimization Algorithm: Epileptic Seizure Detection by EEG Signal

To monitor epilepsy to regenerate a close-loop brain, the Electroencephalogram (EEG) signal is extensively used.Several existing approaches are formulated to identify the seizures, which are based upon the visually evaluating EEG signals which are exclusive as well as the complex procedure if there is an increase in amounts of the channel.A novel approach called Improved Chicken Swarm Optimization Algorithm (ICSOA) is developed and is exploited to train the deepstacked autoencoder (Deep SAE) for identifying epileptic seizures.At first, using the bandpass filter, EEG signals are subjected to pre-processing phase in that noise is evaded.Subsequently, the noise-evaded signals are given as input whereas EEG is to train the Deep SAE that is partitioned into several channels and that channel carries out the feature extraction.Moreover, the features namely Holoentropy, relative energy, fluctuation index, Taylor-based delta AMS, spectral features, tonal power ratio, as well as Linear Prediction Coefficient (LPC) are obtained from each channel.To reduce the dimensionality of the feature, the Probabilistic Principal Component Analysis (PPCA) is used.For the recognition of epileptic seizures, the attained feature vectors are subjected to Deep SAE.Deep SAE training is done by exploiting ICSOA.Therefore, the outcome produced from adopted ICSOA -based Deep SAE is used to recognize the seizures from EEG.The adopted ICSOA -based deep SAE reveals the maximum accuracy, sensitivity, and specificity.