Boehme, Christoph (2011)
Ph.D. thesis, University of Birmingham.
In this thesis a connectionist model for affordance-guided selective attention for action is presented. The selective attention for action model (SAAM) is aimed at modelling the direct route from vision to action using Gibson's affordances to describe actions. In the model complex affordances are encoded in grasp postures. It is shown that this is a viable method to realise aordances. By deriving complex affordances from grasp postures which in turn are derived from grasp affordances and invariants in the visual information, the model implements a hierarchical structure of affordances as postulated in Gibson's affordance concept. This is a novel way of implementing affordances in a computational model. Three studies were conducted to explore the model. The grasp postures generated by SAAM are verified in Study 1 by comparing them with human grasp postures. These grasps were collected in an experiment which used the same stimulus shapes as the model. In Study 2 the attentional behaviour of SAAM is investigated. It is shown that the model is able to select only one of the objects in visual inputs showing multiple objects as well as selecting the appropriate action category (described by a grasp type). Furthermore, it is shown that the bottom-up selection of the model can be influenced by top-down feedback of action intentions. In Study 3 the fndings of Study 2 are applied to stimuli showing two hand-tools (pliers and hammer) to clarify how the findings of Study 2 link to complex affordances and action intentions. The results demonstrate that SAAM offers a novel and powerful way to implement complex affordances. Possibilities for further experimentation and extensions of the model are discussed.
This unpublished thesis/dissertation is copyright of the author and/or third parties. The intellectual property rights of the author or third parties in respect of this work are as defined by The Copyright Designs and Patents Act 1988 or as modified by any successor legislation. Any use made of information contained in this thesis/dissertation must be in accordance with that legislation and must be properly acknowledged. Further distribution or reproduction in any format is prohibited without the permission of the copyright holder.
Repository Staff Only: item control page