Purchase this article with an account.
Michiteru Kitazaki, Yasuyuki Inoue; Perception of human body poses: view dependency and search efficiency. Journal of Vision 2004;4(8):233. doi: 10.1167/4.8.233.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Human body is a very familiar object for us, and some interesting perceptual phenomenon has been reported. Biological motion (Johansson, Scientific American, 1975) and Body-inversion effect (Reed et al, Psychological Science, 2003) suggest that representation of human body is processed configurally and differently from other objects. We investigated view dependency and search efficiency of human body using natural poses and impossible poses. We made 30 natural (possible) poses of 3-D human-body models using Poser 4.0 software, and made 30 impossible poses by bending left or right elbow and knee of the possible poses into biomechanically impossible positions. In the first experiment, a pair of images of poses from 8 different viewpoints was sequentially presented and 10 subjects judged if images were of the same pose or the different poses. In results, the recognition performance across viewpoints decreased as viewpoints differed (view dependency). The view-dependency was less in natural poses than impossible poses. It was, however, not found with inverted pose images. In the second experiment, 8 subjects performed visual search task. When the target was a natural pose, the distracters were various impossible poses, and vice versa. Half of trials were with target, and the others were without target (2, 4, 8, 16 items). In results, we found search asymmetry between natural and impossible poses; the search for the impossible-pose target among the natural-pose distracters (118ms/item with target, 231ms/item without target) was more efficient than the opposite search (150ms/item with target, 312ms/item without target). Taken together, we found natural human body poses were encoded and processed more view-independently and canonically than biomechanically impossible poses. This supports the idea that the representation of human body poses is configurally processed, and suggests that this special processing is constrained by biomechanical relationship of body parts.
This PDF is available to Subscribers Only