Applicathln of Artificial Intelligence Property • on Neural Networks For Differentiation Between Various Electrical Signals

Abstract

In the present work, three groups of humans were investigated, the control group which includes 30 healthy persons, the second group includes 30 epileptic patients (they suffer from brain disorder and convulsions), and the third group includes 10 patients suffer from migrane (headache on one side of the brain). EEG (it is the measure of cerebral electrical activity) signals were very important for the diagnosis of Epilepsy and Migrane, in Epilepsy EEG signals showed sharp spikes and slow rhythm activity while in Migrane EEG signals showed irritative lesion.

Model coefficients represent statistical features of the brain activity and perform data reduction. Moreover these coefficients would be used to estimate the spectrum of the brain activity. Such data reduction is useful when making comparison between brain activity.

Some statistical parameters such as skewness, kurtosis and variation coefficient were calculated for the three groups. Autoregressive (AR)modeling and autoregressive-moving average modeling (ARMA) were used to simulate the brain activity (EEG) to investigate which of the two models is better for simulating the system. It was found that AR model is efficient to simulate the brain activity, because of its stability