Modeling and Reasoning of Contexts in Smart Spaces

Context can be defined to be a meaningful and descriptive state of a physical space, in which relevant activities of the humans and movements of objects in the same space are consecutively performed. Context aware computing thus requires the instantiation and use of a suitable context model and algorithms capable of discerning a most relevant context at the present time. In our previous research, we developed a method based on stochastic analysis, for inferring contexts from sensor datasets collected from a given smart space. By utilizing three statistical techniques including conditional probability table (CPT), K-means clustering, and Principal Component Analysis (PCA), our inference method is able to predict and generate contexts which can potentially happen. Generated contexts are then used as references when the present context needs to be found. In this paper, we build on our prior work and propose a method that analyzes the current state space and determines the present context guided by the collection of generated potential contexts. We first reconsolidate our context models and context graph in which contexts are structured, and then introduce reasoning methods which apply Euclidean distance and Cosine Similarity. PCA is also used to statistically analyze the state space, which helps to achieve better performance. Using experimental evaluation, we validated the accuracy of our proposed approach in identifying the present context.