An Application Based on Artificial Neural Network for Determining Viewpoint Coordinates on a Screen

  • Bulent Turan
  • Halil Ibrahim Eskikurt
  • Mehmet Serhat Can
Keywords: EyeGaze, eye tracker, artificial neural networks, feature extraction.


This study used two different Artificial Neural Networks (ANN) to determine the point on a computer screen that the user is looking at. First, an ANN, called ANN1 was developed to identify the eye region of a laptop user from a webcam image. The computer screen was then divided into 57 × 32 blocks of 24 × 24 pixels. One hundred of these were randomly selected, and 20 images were taken by the integrated webcam while the user was looking at each point. The eye region was found on each image by ANN1. This eye region data was used to train another ANN, called ANN2. Twenty blocks were selected, and 20 different images were used as the test set. The coordinates of the block at which the user was looking were determined by ANN2. The deviations between the actual location coordinates and the location coordinates estimated by ANN2 were small. We conclude that our ANN2 was successfully trained to find the viewpoint of the user.