Action Recognition by Motion Detection in Posture Space

Abstract

The visual recognition of action can be obtained from the change of body posture over time. Even for point-light stimuli in which the body posture is conveyed by only a few light points, biological motion can be perceived from posture sequence analysis. We present and analyze a formal model of how action recognition may be computed and represented in the brain. This model assumes that motion energy detectors similar to those well-established for the luminance-based motion of objects in space are applied to a cortical representation of body posture. Similar to the spatio-temporal receptive fields of regular motion detectors, these body motion detectors attain receptive fields in a posture-time space. We describe the properties of these receptive fields and compare them with properties of body-sensitive neurons found in the superior temporal sulcus of macaque monkeys. We consider tuning properties for 3D views of static and moving bodies. Our simulations show that key properties of action representation in the STS can directly be explained from the properties of natural action stimuli. Our model also suggests an explanation for the phenomenon of implied motion, the perceptual appearance, and neural activation of motion from static images.

Translation and Articulation in Biological Motion Perception

“…The question of whether calculation of translation plays a role in articulation analysis was addressed by two articulation tasks. Our findings of increasing discriminability with longer stimulus duration in both experiments agree with our assumptions and the model of Theusner et al (2014) concerning the temporal evolution of articulation discrimination on neural level. Moreover, both experiments reveal an impact of the translational component on the perception of articulation direction.…”

“…Whereas early findings emphasized local motion signals (Cutting, 1981;Johansson, 1973;Webb & Aggarwal, 1982), results of recent psychophysical studies (Beintema, Georg, & Lappe, 2006;Lu, 2010;McKay, Simmons, McAleer, & Pollick, 2009;Reid, Brooks, Blair, & van der Zwan, 2009;Theusner, de Lussanet, & Lappe, 2011;Thirkettle, Scott-Samuel, & Benton, 2010) strongly support the approach of a formbased analysis by first computing body posture and then body motion (Beintema & Lappe, 2002;Giese & Poggio, 2003;Lange, Georg, & Lappe, 2006;Theusner, de Lussanet, & Lappe, 2014). Appropriate postureselectivity has been found in the occipital face area (Grossman & Blake, 2002;Michels, Lappe, & Vaina, 2005;Vaina, Solomon, Chowdhury, Sinha, & Belliveau, 2001), the fusiform body area (Michels et al, 2005;Peelen & Downing, 2005;Schwarzlose, Baker, & Kanwisher, 2005) and the extrastriate body area (Downing, Jiang, Shuman, & Kanwisher, 2001).…”

“…Hence, while articulation has a deterministic relationship to translation, facing has a probabilistic relationship to translation. If translation computation plays a role in perception of facing and or articulation direction, this influence could link to several possible neural mechanisms intervening on different processing levels of dynamic form analysis as described by Theusner et al (2014).…”

“…In a motion energy model of human body motion detection in 3D space, analogous to the standard model in luminance-based object motion detection (Adelson & Bergen, 1985;Burr, Ross, & Morrone, 1986;Reichardt, 1957;van Santen & Sperling, 1984;Watson & Ahumada, 1985), Theusner et al (2014) described how the visual system may calculate biological motion and its features as they are used in discrimination tasks of many psychophysical experiments. According to their model, a presented body posture activates postureselective neurons as stored 2D body template representations, each one to the extent of its matching with the presented posture.…”

“…Equally, by maximum pooling of body motion energy for each time step and following summation over stimulus duration, articulation direction is derived. With respect to the temporal evolution of the calculation processes, Theusner et al (2014) indicated that body motion-selective neurons discriminate articulation direction after a latency of about 200 ms, reaching a saturation level after about 400 to 600 ms after stimulus onset. In comparison, response latency to form is suspected to be around 119 ms (Oram & Perrett, 1996).…”