Real-Time Activity Prediction and Recognition in Smart Homes by Formal Concept Analysis

In this paper, we introduce a new knowledge-driven approach based on the Formal Concept Analysis (FCA) to predict and recognize Activities of Daily Living (ADLs) in ubiquitous computing environments, in order to duly provide continuous assistance for residents. The proposed approach constructs an incremental inference engine and achieves progressive deductive reasoning to recognize an unfinished ongoing ADL in real-time. For the purpose of finding out the most probable ongoing activity among possible candidates, we propose an assessment based on the root-mean-square deviation (RMSD) to evaluate the relevance of each intermediate prediction. Besides the on-the-fly recognition mode, our approach also possesses high discrimination in differentiating derived and similar activities. Excellent recognition results (almost 100 %) and high prediction accuracies (more than 70 %) are obtained in the experiments.