Enabling Context-aware Applications In Smart Environments‏

Recognizing human activities is an important feature for the development of contextaware applications that are so fundamental to enabling assistive environments. Only once these applications are able to determine the activities that their inhabitants are performing can they assist the individuals and their special needs. In order to do this, it is necessary to build models that can accurately capture and recognize the observed patterns. Equally important is the need to manage and distribute the information that has been inferred, and to provide Quality of Service (QoS) guarantees so that context-aware applications can react effectively to emergent situations. In this thesis, we explore these two problems of knowledge inference and dissemination. We approach the issue of activity recognition from a new perspective, were our goal is that of mining rules to complement an existing algorithm's capability to recognize activities, and thus improve overall accuracy results. For knowledge dissemination, we propose a framework to facilitate QoS in ontology centered context aware pervasive middleware. We believe our work validates two potential ways in which to overcome some of the existing problems for the proper operation of context-aware applications.