An investigation of algorithms to clean RFID data for activity monitoring of the elderly

Abstract

Radio-Frequency Identification (RFID) system functions as a potentially flexible and low cost tool for both object locating and human activity tracking. In this dissertation, a comparative literature review was employed initially, in order to gain wide background and theoretical evidence to answer our research questions. Experimental investigations were carried out, which were focused on examining and evaluating the effectiveness of RFID performance. Bearing in mind the challenges confronted by the elderly, we developed prototypes in our experiments. In order to improve the data reliability and overall performance of RFID application, this experimental investigation methodology was used within a positivistic paradigm. The research focussed mainly on the development and evaluation of tools to clean the RFID data stream and improve the identification of activity. Based on analysis of experimental results, we examine whether fixed (or static) window cleaning method or Statistical sMoothing for Unreliable RFID data (SMURF) middleware (Jeffery, Garofalakis & Franklin, 2006) is a viable and cost-effective candidate to produce more reliable data for human activity monitoring.