Perceptual Adjustment of Eyeball Rotation and Pupil Size Jitter for Virtual Characters

Abstract

Eye motions constitute an important part of our daily face-to-face interactions. Even subtle details in the eyes’ motions give us clues about a person’s thoughts and emotions. Believable and natural animation of the eyes is therefore crucial when creating appealing virtual characters. In this article, we investigate the perceived naturalness of detailed eye motions, more specifically of jitter of the eyeball rotation and pupil diameter on three virtual characters with differing levels of realism. Participants watched stimuli with six scaling factors from 0 to 1 in increments of 0.2, varying eye rotation and pupil size jitter individually, and they had to indicate if they would like to increase or decrease the level of jitter to make the animation look more natural. Based on participants’ responses, we determine the scaling factors for noise attenuation perceived as most natural for each character when using motion-captured eye motions. We compute the corresponding average jitter amplitudes for the eyeball rotation and pupil size to serve as guidelines for other characters. We find that the amplitudes perceived as most natural depend on the character, with our character with a medium level of realism requiring the largest scaling factors.

Is the avatar scared? Pupil as a perceptual cue

“…Avatar creation : We adopted the highly realistic human avatar from Jörg et al 12 The avatar's eyeball movement was driven by a point in a 2‐D plane through a point constraint to mimic someone watching a screen. We adjusted the plane's size and distance from the character to match the real world eye tracking setup.…”

“…We added an offset to the noise to make the mean pupil diameter the same as condition scary. We scaled the noise to result in an average amplitude of 0.0023, a value used in Jörg et al 12 They generated procedural pupil jitter for two motions: a reading motion and a motion inspired by an eye tracking calibration procedure where viewers looked nine points on a white board without any changes in light. We used the average noise amplitude for the nine point motion as that type of motion is most similar to our stimuli.…”

“…Previous research has focused on making the dynamics of the eyeball realistic, 9 as well as the motions of the eyelids 10,11 . With advances in eye tracking, there is now work on adding subtle pupil dynamics, such as pupil jitter 12,13 and pupil light reflex 14 . While this past research has performed perceptual studies to show that adding realistic dynamics to the eyeballs, eyelids, and pupils improves the perceived realism and naturalness of a virtual avatar, there has been little investigation into how cues in the eyes impact the perceived internal state of a virtual avatar.…”

The Security-Utility Trade-off for Iris Authentication and Eye Animation for Social Virtual Avatars

“…Your task is to respond to several prompts about the animation after each trial.". Based on prior work we evaluate each interaction in terms of truthfulness [77], naturalness [43], attentiveness [25], comfort with the avatar [32], and eye contact [58]. After watching each animation the participant used a mouse to respond to the following prompts, using a five point Likert scale from 'Strongly Disagree' to 'Strongly Agree'(1-5):…”

“…Results from Duchowski et al [23] suggest that data-driven eye animations are perceived as more natural than procedural animations. Jörg et al [43] found that subtle variations in the amplitude of noise within data-driven eye animations influenced how natural the animations were perceived. This suggests that a small amount of spatial noise in the signal may be detected, and have a negative impact on the naturalness of eye animations.…”