Purchase this article with an account.
Jeroen J.A. van Boxtel, Hongjing Lu; Action classification through a new reverse correlation technique: the feet are important, as well as their correlated motions. Journal of Vision 2013;13(9):186. doi: https://doi.org/10.1167/13.9.186.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Humans frequently face situations that require the interpretation of actions of another person. It is still largely unknown what information is used by observers to perform action recognition tasks. We employed a new reverse correlation technique to uncover which information is used in the discrimination of point-light walkers and runners. The actor consisted 13 dots on the main joints, the actor rotated in depth (150deg/s), and presentation time was 1 cycle = ~1 sec. Walker and runner cycle lengths were identical. Each trial, observers viewed a point-light actor of which the joints were individually morphed between walker and runner actions, with randomly assigned weights taken from a normal distribution (mean (SD) = 0.5 (0.25)). Weights falling outside the 0-1 range were redrawn. Subjects classified each trial as a walker or runner. The data were submitted to a logistic regression, and we extracted the beta value (i.e., strength) of each joint in the discrimination process, as well as the correlation matrix (determining the interaction between joints). In independent experiments we found that the weight of the feet significantly influenced the discrimination process. The correlation matrix revealed that the feet are also positively correlated (i.e., observers paid attention to both feet concurrently). Additionally, several positive correlations existed between the upper and lower body, mainly between the wrists/elbows and the feet. Previous research had indicated that the feet are important in determining the facing and walking direction of an actor (Troje & Westhoff, 2006, Curr Biol). Our method confirms the importance of the feet in another task (walking versus running discrimination), and did so using a hypothesis-free reverse correlation technique. A further advantage of this technique is its generalizability to any type of action discrimination task (e.g., boxing vs. walking, or boxing vs. waving), and that it reveals inter-joint interactions.
Meeting abstract presented at VSS 2013
This PDF is available to Subscribers Only