Modelo computacional para la generación de indicaciones en la localización de objetos en entornos virtualesaspectos espaciales y perceptivos

Resumen

This doctoral thesis presents a computational model that simulates the process of giving directions for the location of objects within virtual environments. The model allows the automatic generation of directions in natural language, using spatial and perceptual aspects, with reference objects. In order to assist users in the process of location of objects, this work is supported by three pillars: - The proposal of a computational model of perceptual saliency for 3D objects applicable to Virtual Environments. The computational model of saliency facilitates the evaluation and selection of the best reference object to be used in directions. It is based on the object’s features with the major stimulus for the human visual system; three basic features of 3D objects (i.e. color, size and shape) are evaluated from two perspectives: individual and with respect to the context. - A user model and an explicit representation of virtual world semantics. In order to make the directions for the location of objects as relevant to the user as possible, a model to represent previous education and experience of the user as well as his/her cognitive and perceptual attributes is presented. The perceptual characteristics allude to how the user interprets information through the visual system. The cognitive characteristics reflect somehow the processes of the human brain, i.e. how the user saves and manages information throughout time. Likewise, a semantic representation of the virtual environment is necessary for the generation of directions, in which properties and relations between objects in the virtual environment are captured by an ontology. - The automatic generation of directions to locate objects in natural language, this takes into account reference frames and reference objects. For the selection of the best reference object three criteria were taken into account: (1) The perceptual saliency of objects; (2) The probability that the user remembers the location of the object, and (3) Prior knowledge of the user about the class of object. This computational model was implemented as a software component that can be integrated into any virtual environment in which there is a need to guide users in the location of objects, for example when intelligent pedagogical agents within virtual training environments demand users to interact with objects that previously have to be located by the user. The results obtained through a set of experiments allowed us to collect evidence on the effectiveness and usefulness of the proposal, and to verify that the model is adaptive to user characteristics and to the context or situation.