Epileptic Seizure Detection Using Higher Order Moments on Eeg Signals

Brain disorders include Alzheimer’s disease, learning disability, traumatic brain injury, stroke, emotional disorders, seizures, attention deficit disorder, etc....Seizures occasionally known as a ‘fit’ are most commonly occurring brain disorder among humans. Sometimes, a person may have seizure without aura. It is very tedious and expensive to have a person constantly observe each patient and every EEG being observed. It is also cumbersome to review a 24 hour continuous EEG recording. On this backdrop epileptic seizure detection is an important aspect of long term epilepsy monitoring. Epileptic detection, particularly if it is performed online can be of great assistance in identifying sections of EEG likely to have seizures. Even if false alarms are particularly frequent compared to true detection automatic detection can be very useful. The proposed system is based on time domain analysis of EEG signal. Each channel of both seizure and normal EEG data are divided into frames of 256 samples. Then corresponding to each EEG segment higher order statistical features such as mean, variance, standard deviation, skewness, and kurtosis are calculated. After the feature extraction, classification is done using neural networks. This gives a better accuracy than existing methods.