Perceptual Effects of Inconsistency in Human Animations
Animation retargeting is a method of producing human-like animations for use in research, video games, and movies. The procedure consists of capturing the motion of a performer and applying it to a computerized avatar. In practice, the body shape of the avatar can be very different from the body shape of the original performer. Such animations are called inconsistent because they are generated from mismatching shape and motion components. However, actions seen in day-to-day life are typically composed of correlated and consistent information. As a result of this exposure it is believed that the visual system builds up expectations regarding the way a person should look and move. Here I asked whether the human visual system has a special sensitivity to shape-motion inconsistency that would modify the perception of realistic animations. I recorded 10 male performers with either light or heavy body weight carrying out three actions: pushing, lifting, and throwing objects. I extracted shape and motion estimates from each performer, creating consistent and inconsistent animations whose shape and motion components were either from the same performer or combined from performers in different body weight groups. I conducted three experiments in an immersive virtual reality environment. First, when tasked with detecting shape-motion inconsistency, participants had a small ability to select the inconsistent stimuli instead of the consistent stimuli over chance levels. Next, using judgements of animation characteristics as an alternative measure of sensitivity, I assessed the effect of inconsistency on perceived eeriness, humanness, and attractiveness of consistent and inconsistent stimuli. None of these measures were affected by inconsistency. However, in a final experiment where I asked performers to rate the perceived weight and thrown distance of objects, I found an influence of shape-motion inconsistency on the perception of action outcomes. This indicates that the visual system relies on its knowledge of shape and motion to perceive realistic human animations. I propose that animated avatars that involve object manipulations may present an opportunity for the visual system to reinterpret inconsistency as a change in the dynamics of an object, rather than as an unexpected combination of body shape and body motion.