Epileptic seizure detection using non linear analysis of EEG

EEG contains immense information about the brain activity which cannot be understood completely by visual inspection. Powerful signal processing algorithms in EEG analysis can greatly assist the physicians and neurologists to extract such hidden information. It has been found that EEG being a time-varying and non-stationary signal, can be analyzed by non-linear methods. In this paper we tried to evaluate the non linear features like Approximate Entropy, Sample Entropy and Hurst Exponent in epileptic and normal EEG and has obtained clear discrimination between them. These features extracted during interictal phase EEG are potential parameters for exploring the possibility of epilepsy diagnosis from interictal phase EEG. SVM classifier was implemented for the non linear features extracted from EEG. Classification parameters of classifier based on non linear features were calculated.