Inferring Context

Context Awareness is the task of inferring contextual data acquired through sensors present in the environment. ‘Context’ encompasses all knowledge bounded by a scope and includes attributes of machines and users. A general context aware system is composed of context gathering and context inference modules. This paper proposes a Context Inference Engine (CiE) that classifies the current context as one of several recorded context activities. The engine follows a distance measure based classification approach with standard deviation based ranks to identify likely activities. The paper presents the algorithm and some the results of the context classification process. DOI: 10.4018/japuc.2012070102 14 International Journal of Advanced Pervasive and Ubiquitous Computing, 4(3), 13-41, July-September 2012 Copyright © 2012, IGI Global. Copying or distributing in print or electronic forms without written permission of IGI Global is prohibited. Falcao, & Gibbons, 1992; Cheverst, Mitchell, & Davies, 1999; Abowd, Atkenson, Hong, Long, Kooper, & Pinkerton, 1997; Román, Hess, Cerqueira, Campbell, & Nahrstedt, 2002; Hofer, Pichler, Leonhartsberger, Altmann, & Werner, 2002; Wang, 2004; de Deugd, Carroll, Kelly, Millett, & Ricker, 2006; Dey & Abowd, 1999; Riaz, Kiani, Lee, Han, & Lee, 2005). These first generation systems lacked the inference support based on classification techniques and were primarily RBA. Many researchers have identified RBA as a promising technique due to its efficiency in computation (Chen, Finin, & Joshi, 2003; Mahmud, Iltaf, Rehman, & Kamran, 2007; Gu, Pung, & Zhang, 2004; Fahy & Clarke, 2004; Guo, Gao, Ma, Li, & Huang, 2008). Use of context matching (Samulowitz, Michahelles, & Linnhoff-Popien, 2001) has provided an improvement on RBA. The design of a context aware system following a rule based technique though considered efficient lacks flexibility in terms of learning new rules and situations. Moreover, the strain on the developer to introduce new rules and modify them in different software engineering phases is high. Context aware systems are thus restricted to the quantity of the rules developed by the engineer. To provide learning capability in the context aware system, some researchers have proposed context recognition processing as the alternative to the inference task (Korpipaa, Mantyjarvi, Kela, Keranen, & Malm, 2003; Mayrhofer, 2004; Blum, 2005; Brdiczka, Crowley, & Reignier, 2007; Yuan & Wu, 2008). The context aware system recognizes the current situation and proposes appropriate actions as suggested by the developer. Use of supervised learning has provided encouraging results in context recognition (Korpipaa, Mantyjarvi, Kela, Keranen, & Malm, 2003; Brdiczka, Crowley, & Reignier, 2007). The researchers following the context recognition generally agree on the distance being the criterion for similarity among different contextual situations (Mayrhofer, 2004; Brdiczka, Crowley, & Reignier, 2007; Yuan & Wu, 2008). The similarities in context highlight the importance of contextual conflicts that may arise in dynamic situations. Priorities have been proposed as a general remedy to resolve conflicts (Mayrhofer, 2004; Shin & Woo, 2005).