Analysis and classification of eeg signals for seizure detection using novel soft computing strategies

Abstract

Electroencephalography is one of the most important medical tools for evaluation and treatment against neurophysiologic disorders related to epilepsy. eeg can provide improved insights with careful analysis of valuations and valuable mechanisms that cause epileptic disorders. hence new research requires the understanding of the mechanisms for creating epileptic disorders. the detection of epileptic discharges occurring in the eeg between seizures is essential in epilepsy diagnosis. several automatic eeg signaling classifications and seizure discovery methods have been used in recent years, using various approaches. still, most of the existing techniques suffer from missed detection, have a high rate of false alarms and low accuracy. herefore, this research work provides eeg, an automated mechanism for improving epilepsy seizure detection, using multiple domain features and non-linear analysis to detect the sequencing process efficiency based on three advanced novel soft computing strategies, for example, principle component analysis (pca) with template matching, nonlinear vector decomposed neural network (nvdn) and asymmetrical back propagation neural network (abpn). improving the accuracy of the proposed system is primarily based on designing an appropriate representation space. identifying the combination of all the extracted features that increase the isolation between classes and classifiers can accurately categorize the capture signal. in this work, the physionet dataset eeg records signals have been used to test the algorithm efficiency, and the results are possible to implement an automatic seizure diagnostic system. the proposed pca with template matching solution is to automate and introduce the eeg signal classification approach. the proposed method is used to classify eeg signals into two classes. in the proposed way, the eeg will begin to extract the features using the decomposition of signals sub-columns to disassemble signals. these features, derived from the details and dwt companions are approxim