Classification of seizure and non-seizure activity in seizure patients using time-frequency domain processing of gamma band EEG signals

In this paper, time-frequency domain operation exploiting wavelet-EMD analysis is performed on the gamma band (40-80 Hz) oscillations of EEG signals for the purpose of classifying seizure (ictal) and seizure-free (inter-ictal) activity in seizure patients. Dominant Intrinsic Mode Functions (IMFs) resulting from wavelet-EMD operation are utilized to propose higher order statistics, entropy and root mean square (r.m.s) based features. One way ANOVA test has been performed for selecting r.m.s value as the most effective feature for the discrimination of ictal and inter-ictal stages in seizure patients using k-Nearest Neighbor classifier. Comprehensive simulations are passed with a standard EEG dataset. It is found that the suggested scheme is accomplished with causing greater accuracy, sensitivity, and specificity in comparison to that acquired by using an up to date method engaging the identical EEG dataset.