A new qualitative spatial recognition model based on Egenhofer topological approach using C4.5 algorithm: experiment and results

Abstract

Ambient technologies and ubiquitous computing constitute together an emerging trend of research bringing new possible solutions to many problems of human life. One of them is the technological assistance of the elders suffering from cognitive deficit with their everyday life activities inside what is called a smart home. The main issue in implementing such technology is the recognition of the activities of the resident. This problem consists in inferring the minimal set of possible ongoing activities using models defined in a plans library. To achieve that, most works propose to exploit different types of constraints (logical, temporal, etc.) in order to eliminate a maximum of incoherent hypotheses. However, very few works considered exploiting the spatial aspect related to the movement of objects and to their relations in space. In this paper, we propose to add a spatial pre-filter based on a topological approach from Egenhofer to discriminate implausible ongoing activities before applying a C4.5 decision tree to choose from the remaining hypotheses. Furthermore, this paper presents promising results we obtained from an experiment on that model using real case scenarios built from clinical trials that we conducted with Alzheimer's patients.