Measuring gaze depth with an eye tracker during stereoscopic display

While determining 2D gaze position using eye tracking is common practice, the efficacy of using eye tracking to measure 3D gaze point in a stereoscopic display has not been carefully studied. In this paper we explore this issue using a custom Wheatstone stereoscope augmented with an eye tracker. In a pilot study, we showed that there is strong evidence that eye vergence measurements do, in fact, track the depth component of the 3D stereo gaze point. In a subsequent full study, we compare depth estimation for a scene viewed in four different ways: without either stereo or motion parallax, without stereo but with motion parallax, with stereo but without motion parallax, and with both stereo and motion parallax. We show a significant effect related to depth from eye vergence for both of the stereo cases, but no effect for the monoscopic cases. Since depth from vergence is very noisy, we use a Butterworth filter to smooth the data and show that this greatly improves the stability of the depth reading without significant lag or loss in accuracy. We also demonstrate that using quadratic regression to perform a depth calibration can greatly improve the accuracy of the depth measurement.