Eyegaze Tracking In Handheld Devices

摘要

Detection of eye gaze helps in analyzing human behavior and personalization of interaction with multimedia devices. Accurate localization of fixations and analysis of saccades have been the focus of research in this area. This paper presents a robust methodology for gaze detection, in unobtrusive manner using low power hand-held device. The front camera is used for capturing facial video of the user and gaze detection is done through mapping of iris center points to screen co-ordinates. The major challenge has been in reducing the noise in the eye region of the captured video. Initially, as part of the training phase, the screen is partitioned into grids and a model is generated for the noise distribution of the gaze with respect to the fixation on the screen. This model is used during the test phase to detect the fixations on a particular image. Experimental results indicate that the overall gaze detection accuracy on a particular grid is approximately 83% and on a test image is 100%.