Data fusion for wireless sensor networks: an assisted living case-study
Michael A. Collett, M. Salter, D. Adamson · 2009
Wireless sensor networks offer many new capabilities for distributed, ubiquitous sensing, which are well suited to in-home care monitoring. A key challenge is generating meaningful information from the data gathered. This paper presents an analysis of data gathered from a home-monitoring testbed. Firstly a data-only approach was taken by analysing the data for frequently occurring patterns. Secondly, prior knowledge of the system was employed to predict the sensor patterns which would result from likely behaviour of the volunteer. It was found that a data-only approach allows analysis to be generalised to many systems and applications, but incurs the risk that rare but important events may not be identified directly. We also generated system performance metrics, finding that the testbed has a 97% probability of accurately logging one specific simple three-part activity. (4 pages)