Modeling of Activities as Multivariable Problems in Smart Homes Case Study - Recognition of Simultaneous Activities

摘要

Resident of a smart home, who may be an Alzheimer patient needing permanent assistance behaves non-linearly to achieve the intended goals. He actuates the world by realizing actions, activities which can be observed through the embedded sensors of the smart home. To assist him automatically and live independently at home, it is needed to capture information and knowledge from world to reason if the world state is normal and to evaluate how much the intelligent system succeeds; therefore, for recognizing the activities and their correct realizations, we propose to consider the activities as a sort of fuzzy temporal concepts that can be formalized as a multivariable function. Perceiving the world, an Activity Recognition System makes hypotheses and concepts about the observations. These hypotheses are resumed in a smoothing line and at the recognition time, the activities functions check how much the observations are close to their smoothing line. Finally, the activities are ranked based on the inferred similarities to the observations. All the introduced processes are data-driven and a case study that deals with recognition of simultaneous activities based on the proposed modeling approach is presented.