Modelling correlations for Body Sensor Network information

Abstract

Body Sensor Networks (BSNs) are the natural candidates to provide multi-parameter patient monitoring. Tapping into multiple inputs and correlating them to infer new information, cleaning received data and inferring state should be objectives of BSNs. These systems will need to deduce information from a variety of raw sensor data and accuracy in their results will be paramount. Apart from having several different types of sensors (producing different types of data), BSNs will also have several applications wanting to access information. Not all of the information will be directly sensed, but some can be inferred from the raw sensor data. We propose a framework that enables modularization of information and its correlation. This enables re-use by different applications and optimization of the collection and calculation of the requested information by the system (the BSN). The framework also allows defining dependencies between modules for information production. Our architecture provides an abstraction on the way information is assessed and its processing flow. Applications issue requests to the middleware with requirements to be met. So we will discuss the optimization of resources, while honouring requirements.