The accurate location of epileptic seizures by interpreting an EEG (Electroencephalogram) signal is highly demanding and involves skilled neurologists. In this study, the EEG is analyzed using the Tunable-Q Wavelet Transform (TQWT) method for identifying seizures by splitting an EEG signal into several sub-bands. The entropy computed for each sub-band signifies the nonlinearity in an EEG signal. The other novel parameters viz correntropy, centered correntropy (CCE) and correntropy coefficient assess the nonlinearity of EEG signal and forms the basis for classification. The study has been done on the freely accessible Bonn University EEG database and outperforms in terms of complexity. When contrasted to the existing state-of-the-art methods, 100% accuracy has been achieved in discriminating seizure, seizure-free signals, and non-seizure EEG signals using Random Forest Classifier. Moreover, the computation of the proposed features is fast, and the system is easy to implement.