EM Based Non-Linear Regression and Singular Value Decomposition for Epilepsy Classification

One of the major disorders affecting the human brain is epilepsy and it impairs the daily lives of the patients. Epilepsy is caused by sudden and random incidence of seizures. One of the most powerful tool for diagnosing various neurological disorders including epilepsy is Electroencephalogram (EEG). Recognizing the people who have a disorder of the brain by means of inspection of EEG signals visually is a very tough task to perform. The discrimination of EEG signals by visual inspection consumes a lot of time, error prone and pretty expensive process and it is not sufficient enough for obtaining reliable information. Better diagnostic methods for analyzing brain-related disorders, especially in case of epilepsy can be done by the analysis and classification of EEG signals. As the recordings of the EEG signals are long, Singular Value Decomposition (SVD) is used to reduce the dimensions of the EEG data. The dimensionally reduced values are then post classified with the aid of Expectation Maximization Dependent Non Linear Regression for epilepsy classification from EEG signals. The results reported an average accuracy as of 94.90%, average quality value as of 19.33, average time delay as of 1.84 and average Performance Index as of 88.97% is obtained.