Imperceptible Gaze Guidance Through Ocularity in Virtual Reality

TL;DR

This work investigates imperceptible gaze guidance in VR by exploiting ocularity, the differential input between the two eyes, with $O \in \{-1,0,1\}$. Grounded in the V1 Saliency Hypothesis, the authors test whether ocularity singletons can attract or misdirect gaze in an odd-one-out search task using 200 Hz eye-tracking in a VR headset. Results show faster detection and focused gaze when the target itself is an ocularity singleton, while a distractor singleton can hinder performance, supporting a bottom-up, V1-driven mechanism that operates without perceptible scene changes. These findings suggest practical, depth-preserving gaze guidance for immersive XR applications, with broad implications for education, training, gaming, and safety scenarios, alongside important ethical considerations and privacy safeguards.

Abstract

We introduce to VR a novel imperceptible gaze guidance technique from a recent discovery that human gaze can be attracted to a cue that contrasts from the background in its perceptually non-distinctive ocularity, defined as the relative difference between inputs to the two eyes. This cue pops out in the saliency map in the primary visual cortex without being overtly visible. We tested this method in an odd-one-out visual search task using eye tracking with 15 participants in VR. When the target was rendered as an ocularity singleton, participants' gaze was drawn to the target faster. Conversely, when a background object served as the ocularity singleton, it distracted gaze from the target. Since ocularity is nearly imperceptible, our method maintains user immersion while guiding attention without noticeable scene alterations and can render object's depth in 3D scenes, creating new possibilities for immersive user experience across diverse VR applications.

Figures (5)

• Figure 1: We conducted an odd-one-out visual search task where the participant is asked to find the orientation singleton (a unique bar facing a different direction among uniformly oriented bars). In this case, a random distractor is rendered as an ocularity feature singleton which causes the participant to instinctively look at the distractor instead of finding the target. Ocularity features deal with individual ocular inputs to each eye (e.g. present in one eye while invisible to another). Such an alteration is not perceivable to the beholder as they only see the combined image from both eyes
• Figure 2: A. Middle: The game environment on Unity showing the Player facing the stimulus. Specifically, the stimulus mimicks a 2D grid composed of 3D objects floating in space. Head rotation is disabled to simplify the eye tracking data collection process. The gaze vector is provided by the eye tracking module and is translated to 2D screen coordinates (XY) in pixels. The player's head movement in-game was disabled. Left and Right: The game as seen from the left and right displays, respectively. In this example, a trial with a BAM condition is shown where non-targets are displayed to the right eye while the target (encircled in red in the three images) is shown to both eyes. B. Different configurations in ocularity used in the task. Each image pair represent a simplified example of what the participant sees in VR through the left and right eye display. In this example, all the non-targets are displayed to the right eye. In the experiment, this is randomly selected. Similarly, the orientation of the bars are randomly selected. Depending on the condition, the ocularity singleton can either be a target or a distractor.
• Figure 3: Mean trial duration, accuracy and eye tracking results across different conditions.
• Figure 4: Contrastive search behavior as exemplified by sample scanpaths taken from BAM (left) and BAMI (right): the search ended quickly for salient target conditions with the first fixation landing usually on the target (star) itself while the participant produced more dispersed fixations in the presence of salient distractors (circle) where the distractor itself is the first to be gazed at by the subject. The scanpath color indicates temporal order, with darker shades representing earlier fixations.
• Figure 5: A comparison of BAM/I, DC/I, MAB/I, and BASE conditions in terms of ocular contrast where each item in the grid is represented by its corresponding ocularity $O$. Each condition shows a simplified $4\times4$ grid where the ocularity singleton (in all conditions apart from BASE) is found on the 3rd column in row 2. BAM/I and DC/DI conditions showed significant effect on the participant's search behavior while MAB/I demonstrated a small change in search behavior.