AUTHOR's DRAFT Design and Implementation of Gaze Tracking Headgear for Nvidia 3D Vision R  Sunu Wibirama and Kazuhiko Hamamoto * , Member, IEEE Abstract— The usage of Nvidia 3D Vision R  is increasing rapidly, ranging from gaming to research purposes. However, researchers in human computer interaction and virtual reality are constrained by hardware configuration since current com- mercial gaze tracking systems are not specifically designed to be used with Nvidia 3D Vision R  . In this paper, we present a novel prototype of gaze tracking headgear which can be used appropriately with Nvidia 3D Vision R  . We explain design consideration and detail implementation of our gaze tracking headgear. We also evaluate our gaze tracking system by measuring gaze accuracy on stereoscopic display. Experimental result shows that the average gaze estimation error is less than one degree visual angle. Index Terms— gaze tracking, active 3D technology, virtual reality, Nvidia 3D Vision R  I. INTRODUCTION Virtual environment (VE) has been widely used for enter- tainment, gaming system, telepresence, and simulating real or imagined scenario, such as design and training processes in industrial and collaborative task [1]–[3]. Stereoscopic display technology is commonly used to enhance user ex- perience in three-dimensional (3D) VE. Stereoscopic display technology is developed based on stereopsis in human visual system. Since human eyes are separated horizontally, each eye has its own view of the world scene. Consequently, both eyes receive slightly different images. To achieve 3D experience, human brain has to perceive left and right images as a single image. Using information of screen disparity, which is the distance of corresponding point in the left and right images, the depth information can be extracted. Fig.1 shows a brief concept of stereoscopic viewing. Recently, one of active 3D technology that gains wide attention is Nvidia 3D Vision R  . Nvidia 3D Vision R  system consists of two main hardwares: infra-red (IR) emitter and a lightweight active shutter glasses (50 gram of weight), as depicted in Fig.2. Nvidia 3D Vision R  is widely used in various applications since it preserves color compared with anaglyph-based stereoscopic image. There are many applications that utilize Nvidia 3D Vision R  , including 3D haptic-based modeling system [4], augmented reality with freehand interaction [5], navigation for visually impaired people [6], digital exhibition of archaeological structure [7], This research is supported by AUN/Seed-Net, JICA, and Tokai University under long-term research grant ID J1110112 ∗ Sunu Wibirama and Kazuhiko Hamamoto are with Graduate School of Science and Technology, Tokai University, Tokyo, Japan 108-8619. Email: sunu@jteti.gadjahmada.edu, hama@keyaki.cc.u-tokai.ac.jp Fig. 1. Concept of stereoscopic viewing. By fusing left and right images, the brain perceives virtual 3D object. Parallax angle is defined as the difference between angle formed by both eyes-virtual object and both eyes- display plane (θ-α). Fig. 2. Nvidia 3D Vision R  system: infra red (IR) emitter (left panel) and active shutter glasses (right panel). The shutter glasses is synchronized with 3D display during stereoscopic images exposure with 120 Hz of shutter frequency. welding training for industrial worker [8], and performance investigation for sport player [9]. On the other hand, research on gaze tracking as human- computer interface and visual search tools in 3D space is emerging. Gaze tracker, compared with other input devices such as mouse or keyboard, has been a choice to investigate user attention and visual perception in VE. Moreover, as an interactive interface, human gaze is considered faster than speech or gestures. Essig et al. [10] used a commercial head-mounted gaze tracker to estimate position of stimulus presented in anaglyph stereogram. Lee et al. [11] suggested the usage of monocular gaze tracker to measure user gaze in 3D space. Daugherty et al. [12] used remote gaze tracker to compute point of gaze over anaglyphic stereogram video while the user has to wear blue and red filter to see the 3D video. Due to the nature of common gaze tracker which is not specifically designed for active 3D technology, the usage of Nvidia 3D Vision R  system is normally excluded in previous research works. In case the user has to wear Nvidia 3D Vision R  glasses, an appropriate gaze tracking headgear should be developed.