In this paper, we propose a new set of efficient features for the gaze estimation system that successfully works with a simple and cheap eye gazing system. These features are composed of Pupil – Glint vectors, Pupil – inner-eye-Corner (canthus) vectors, Glint – inner-eye-Corner vectors, distance vector that connects 2 inner-eye-corners, angles between Pupil – Glint vectors and the Glint – inner-eye-Corner vectors, and deviation angle. These features together can capture the head movement and positions of the eyes relative to the position of the screen effectively. From the investigating results, ANN with 2 hidden layers provided the best classification in 15 regions of interest under the proposed features which has an accuracy of 97.71%, delivering better performance than the other techniques. Our proposed features are applicable for real-time applications and fit with low cost and simple system, therefore, it can be widely used for disabled and special need persons in various HCI applications.