Reconhecimento de atividades em casas inteligentes: uma abordagem não intrusiva explorando processamento semântico

Abstract

In recent years, the techniques of Activity Recognition have attracted increasing attention. Among the many applications, special interest lies in the ubiquitous domain of e-Health, where automatic activity recognition is used in rehabilitation systems, chronic disease management, elderly monitoring, as well as in personal wellness applications. This interest has been motivated by the aging of the population. This aging results in significant socioeconomic challenges in the public health sector, as well as in the incidence of chronic-degenerative diseases, with dementia being one of the most worrying. An alternative that has been widely proposed is the use of intelligent houses (ambient assisted living), in which the residences of people under treatment should contemplate computational services that can assist them in their daily practices, in the most transparent way possible. This work has the acronym EXEHDA-AR (EXEHDA-Activity Recognition), and its main objective is to contribute with the Context Recognition and Adaptation Subsystem of the EXEHDA (Execution Environment for Highly Distributed Applications) middleware, enabling it to meet the demands Of Activity Recognition, using a Context Awareness approach. The EXEHDA-AR explores semantic processing to provide Activity Recognition in smart home, for which we have proposed components to be integrated with the EXEHDA middleware as well as an ontological model. The collected context data are grouped according to the concept of time-sliding window. When evaluated the EXEHDA-AR obtained an average accuracy of 94.36% in the Recognition of Activities. These results indicate that methods based on semantic processing constitute a viable alternative for the Recognition of Activities with low level of intrusion, indicating the continuity of research efforts.