Research Article

Automatic Epileptic Seizure Detection Using PSO-Based Feature Selection and Multilevel Spectral Analysis for EEG Signals

Table 2

The list of features.

TypeFeature name

Time domain (31)Maximum, minimum, mean, variance, line length, skewness, kurtosis, nonlinear energy, root mean square amplitude, the variances of the first and second difference signal, Hjorth parameters (activity, mobility, and complexity), zero crossing (the number of zero crossings of original signal and its first and second derivatives), number of maxima and minima, wave form factor, peak factor, pulse factor, margin factor, errors of AR modeling with order 1-9
Frequency domain (10)Total power, peak frequency, median frequency, center frequency, frequency variance, root mean square frequency, wavelet energy, spectral edge frequencies (3)
Information theory (5)Spectrum entropy, singular value entropy, sample entropy, wavelet entropy, energy entropy