DETECT FEATURES OF SEIZURE SERIES BY EXPONENT FUNCTION

Abstract

In medical diagnostics, epilepsy is recognized by observation of EEG. Epilepsy is often highly correlated with a series of seizures that last longer 3 seconds. Detection of these features is the most important problem for automatic recognition purposes. In signal processing, the recognition accuracy often depends on the feature extraction. Recently, deep learning was a useful tool for the feature extraction of epileptic seizures from EEG. Even though deep learning is the best method for this issue, it requires strong hardware for its computing. This paper will point out that the exponent function can extract the feature of seizures. The exponent function is simple therefore this solution may setup on any platform. The experimental results show that our method has the advantage in case of complicated epileptic seizures. In EEG, the exponent function indicates the location of seizures in a time series. The detection model is extended for epilepsy recognization.