Identifying Epileptic Seizure by Optimized Feature Extraction Process Using the Method of Feature Fusion Technique

Abstract

Epilepsy is a brain disorder that results in seizures; in general seizure is a suddenly occurring uncontrollable electrical disturbance in the brain. These disturbances in the brain can lead to changes in behaviour, feelings, movements, etc. It is highly essential for the patients suffering with epilepsy to be diagnosed and treated. The normal detection of epilepsy is done using EEG signals which are time consuming. This paper aims at proposing a methodology to diagnose epilepsy by the use of EEG signals by establishing a correlation between statistical calculations and EEG signals. A various set of features are applied to the epilepsy and non-epilepsy dataset. Features such as time domain frequency which include mean, skewness, variance, kurtosis, standard deviation, approximate entropy, zero crossings, power spectrum and frequency domain features that include signal energy and total signal area, average DWT coefficient, signal relation features and human brain graph features. Further, considering these features, feature fusion and optimization aka FFO is carried out which helps in analysing the features in an optimal way for further classification. Moreover, feature fusion and its optimization helps in exploring the new features that helps in enhancing the distinguish between classes. These features help diagnosis of the brain disorder in a very time efficient manner with higher accuracy. In this paper, we propose a feature fusion methodology for the most efficient working of the system.