| Paper: | PS-2B.38 | ||
| Session: | Poster Session 2B | ||
| Location: | H Fläche 1.OG | ||
| Session Time: | Sunday, September 15, 17:15 - 20:15 | ||
| Presentation Time: | Sunday, September 15, 17:15 - 20:15 | ||
| Presentation: | Poster | ||
| Publication: | 2019 Conference on Cognitive Computational Neuroscience, 13-16 September 2019, Berlin, Germany | ||
| Paper Title: | A Model of Full-body Kinematics-based Visual Attention in Daily-Life Tasks | ||
| Manuscript: | Click here to view manuscript | ||
| License: |  This work is licensed under a Creative Commons Attribution 3.0 Unported License. |
||
| DOI: | https://doi.org/10.32470/CCN.2019.1373-0 | ||
| Authors: | Alex Harston, Chaiyawan Auepanwiriyakul, Aldo Faisal, Imperial College London, United Kingdom | ||
| Abstract: | Visual attention and motor actions are intrinsically linked and tightly spatiotemporally coupled in real-world behavior, and yet very few studies of natural gaze behavior account for the dynamics of the body, thereby missing a fundamental aspect of the perception-action loop. To address this, we experimentally capture whole body kinematics and time-synced gaze in a natural, high- dimensional task, to investigate the influence of motor actions on gaze behavior. We use a combination of linear and nonlinear autoregressive models with exogenous body input to assess the predictive power of prior gaze and motor dynamics on future gaze location, and find that our nonlinear model significantly outperforms previous linear models for predicting natural gaze dynamics, and that incorporating whole body kinematic information into our model significantly improves gaze prediction performance versus simple gaze autoregression. Incorporating this body information into visual saliency models helps improve our understanding of visuomotor interactions in the real world. | ||