High frequency oscillations (HFOs) have been considered as a promising clinical biomarker of epileptogenic regions in brain. Due to their low amplitude, short duration, and variability in patterns, the visual identification of HFOs in long-term continuous intracranial EEG (iEEG) is cumbersome. The aim of our study is to improve and automatize the detection of HFO patterns by developing analysis tools based on an unsupervised k-means clustering method exploring the time-frequency content of iEEG. The clustering approach successfully isolated HFOs from noise, artifacts, and arbitrary spikes. We tested this technique on three subjects. Using this algorithm we were able to localize the seizure onset area in all of the subjects. The channel with maximum number of HFOs was associated with the seizure onset.