Electroencephalogram (EEG) signal interpretation has been developed for various purposes such as brain health examination, brain detection, brain trauma, emotional condition, and even predict the response that will occur. The complex form of EEG signals will complicate one's interpretation visually so that it requires neurologists to deduce it. One of the brain disorders that are of concern and can be detected through EEG is epilepsy. EEG signal patterns can be identified through excessive brain cell activity before or after a person experiences seizures without cause. In this study, we proposed an EEG epilepsy signal recognition using Wavelet Time Entropy (WTE) as the main modality to obtain signal features. 300 EEG signal consisting of 3 classes (normal, interictal, seizure) has been tested with the highest accuracy result of 86.3% generated by Db 2 with decomposition level 2 or 3 using cubic Support Vector Machine (SVM)