Many classification algorithms have been developed to distinguish brain activity states during different mental tasks. Although these algorithms achieve good results, they require many training loops to make a decision. As the complexity of an algorithm grows, it becomes more and more difficult to execute commands in real time. The detection of eye movement from brain activity data provides a new means of communication and device control for disabled and healthy people. This paper proposes a simple algorithm for offline recognition of four directions of eye movement from electroencephalographic (EEG) signals. A hierarchical classification algorithm is developed using a thresholding method. A strategy without a prior model is used to distinguish the four cardinal directions and a single trial is used to make a decision. Using a visual angle of 5°, the results suggest that EEG signals are feasible and useful for detecting eye movements. The proposed algorithm was efficient in the classification phase with an obtained accuracy of 50-85% for twenty subjects.