The generation of actual sensory data in real-world deployments of pervasive spaces is very costly and requires significant preparation and access to human subjects. This situation can be mitigated if practical forms of sharing of existing datasets are enabled among the research community. In this paper we address two main problems. First, we propose a standard for the representation of smart space datasets, based on a careful examination of several existing data. The standard specification should allow researchers to effortlessly position their existing or future datasets for sharing. We briefly present the specifications. Second, to enable higher utility of shared datasets, we propose algorithms and tools that can extend a shared dataset into a similar set of a slightly customized pervasive space (e.g., an original space with additional sensors/actuators or behaviors). Specifically, we propose the use of machine learning algorithms to generate the additional patterns of events and to automatically integrate them into the original shared dataset.
Mendeley saves you time finding and organizing research
Choose a citation style from the tabs below