Epilepsy is a brain disorder which stems from the abnormal activity of neurons and recording of seizures has primary interest in the evaluation ‎of epileptic patients. A seizure is the phenomenon of rhythmicity discharge from either a ‎local area or the whole brain and the individual behavior usually ‎lasts from seconds to minutes. In this work, empirical wavelet transform (EWT) is applied to ‎decompose signals into Electroencephalography (EEG) rhythms. ‎EEG signals are separated into the delta, theta, alpha, beta and gamma ‎rhythms using EWT.‎ The proposed method has been evaluated by the benchmark dataset which is freely downloadable from the Bonn University website. Ellipse area (A) and shortest distance to 45 and 135-degree lines are computed from the 2D projection of reconstructed phase space (RPS) of rhythms as features. After that, the genetic algorithm is used as feature selection. Finally, selected features are fed to the K-nearest neighbor (KNN) classifier for the detection of the seizure (S) and seizure-free (SF) EEG signals. Our proposed method archived 98.33% accuracy in the classification of S and SF EEG signals with a tenfold cross-validation strategy that is higher than previous techniques. Manuscript profile