Tele Alert System for Epileptic Seizure on a Study of EEG Signal Classification by GBE-NLSVM through ICA Preprocessed and AR Extracted Signal in a BCI System

Abstract

Abstract Brain computer interface is an action of translating the brain signal into a command for activating artificial object such as limb. BCI is the collaboration of biomedical, electrical, computer, and mechanical engineering. An action potential is created in the form of electrical signal in the brain for every action of a human being, either physical or mental. The patient himself suffering from epileptic seizure poses danger severely during the absence of continuous monitoring. Taking care of epileptic patients from remote locations has become essential since the patient loses his whole control during epileptic seizure. This paper presented an epileptic tele alert system (ETAS) for a patient being monitored from out of the hospital premises. The brain signals tapped using a noninvasive electro encephalographic (EEG) electrode was given to independent component analysis (ICA) to preprocess the tapped signal. The auto regressive method (AR) was employed to extract the feature from training the brain signal for the normal and abnormal condition of the patient. The support vector machine technique named Gaussian basis function non-linear support vector machine (GBF-NLSVM) was used to classify the signal that is a vulnerable point in the cause of the epileptic seizure with respect to brain action potential for various statuses of activities. The frequency beyond the beta level was identified and the signal was transformed as a command for activating handheld devices using microcontroller via global system for mobile communication (GSM). The MATLAB, Simulink software having built in functions for studying the brain signal was used to analyze the brain signal. The proposed model discussed the signal tapping, feature extraction, classification, and activation of mobile phone using microcontroller. The proposed system incorporating ICA, AR, and GBF- NLSVM was compared with other techniques for identifying epileptic seizure and ensured that the system provided about 97 % of accuracy over the other standalone technique.