Noncontact Binocular Eye-Gaze Tracking for Point-of-Gaze Estimation in Three Dimensions

Abstract

Binocular eye-gaze tracking can be used to estimate the point-of-gaze (POG) of a subject in real-world 3-D space using the vergence of the eyes. In this paper, a novel noncontact model-based technique for 3-D POG estimation is presented. The noncontact system allows people to select real-world objects in 3-D physical space using their eyes, without the need for head-mounted equipment. Remote 3-D POG estimation may be especially useful for persons with quadriplegia or Amyotrophic Lateral Sclerosis. It would also enable a user to select 3-D points in space generated by 3-D volumetric displays, with potential applications to medical imaging and telesurgery. Using a model-based POG estimation algorithm allows for free head motion and a single stage of calibration. It is shown that an average accuracy of 3.93 cm was achieved over a workspace volume of 30 x 23 x 25 cm (W x H x D) with a maximum latency of 1.5 s due to the digital filtering employed. The users were free to naturally move and reorient their heads while operating the system, within an allowable headspace of 3 cm x 9 cm x 14 cm.

Assessment of accuracy for target detection in 3D-space using eye tracking and computer vision

“…We consider as benchmarks the methods presented in [Abbott and Faisal 2012] which has a workspace of 60 cm x 50 cm x 40 cm located 50 cm away from the subject and in [Hennessey and Lawrence 2009] which uses a workspace of 30 cm x 23 cm x 25 cm about 25 cm away from the subject. As mentioned earlier, to the best of our knowledge, no other study describe the workspace considered while assessing the accuracy of a 3DPOG method, thus leaving no more possibilities for comparison.…”

“…Almost every research group using this method proposes some degree of refinement to the technique to improve accuracy. However, it is worth noting that very few papers [Abbott and Faisal 2012;Hennessey and Lawrence 2009]. describe the workspace considered during the accuracy assessment of their method, thus leaving little possibility for comparison between methods.…”

Multi-depth cross-calibration of remote eye gaze trackers and stereoscopic scene systems

“…In 2009, Hennessey and Lawrence claimed to be first in devising a binocular system for estimating the 3D coordinates of where one is looking in the 3D world, by using vergence [5]. In their experiments, fixated objects were close to test subjects and contained in a 1.725 m 3 volume.…”

“…Hennessey and Lawrence presented a 3D PoG method which employs eye vergence to estimate the 3D position of a fixated object [5]. In their experiments, fixated objects were contained in a 0.01725 m 3 volume located in front of the subject.…”

3D gaze cursor: Continuous calibration and end-point grasp control of robotic actuators

“…The 3D gaze fixating point is inferred from the calculation of the closest intersection point of the aforementioned vectors in the 3D space. The overall determined accuracy of the system was 3.93 cm in Euclidean space [15]. Alternative solutions to solve the insufficient accuracy in the depth dimension have attempted to use 2D gaze control while performing plane selection independently.…”