Qualitative Spatial Activity Recognition Using a Complete Platform Based on Passive RFID Tags: Experimentations and Results K. Bouchard , B. Bouchard, A. Bouzouane LIARA Laboratory, Universite du Quebec a Chicoutimi (UQAC) 555 boul. Universite, Saguenay (QC), Canada, G7H 2B1 {Kevin.Bouchard, Bruno.Bouchard, Abdenour.Bouzouane}@uqac.ca Abstract. Smart home has become a very active topic of research in the past few years. The problem of recognizing activity inside a smart home is one of the biggest challenge researchers have to face in this discipline. Many of them have presented approaches exploiting temporal constraints in order to maximize the efficiency and the precision of their recognition model. However, only few works investigate the spatial aspects characterizing the habitat context. In this paper, we present a new algorithm and a complete experiment showing the importance of taking into account spatial constraints in the recognition process. The goal of this paper is to demonstrate that recognition algorithms will benefit from exploiting spatial constraints. Keywords: Smart Homes, RFID, Activity Recognition, Spatial Constraints. 1 Introduction Smart home technologies have become an interesting and very active research trend [1], bringing hope in the effort to postpone the institutionalization of the elders. A smart home can be seen as an augmented environment with miniaturized processors, multi-modal sensors that are embedded in common objects, and software agents communicating between each other [2]. Those environments must take decisions, while remaining non-intrusive, in order to help the resident completing his tasks. The first fundamental step in smart home assistance is to be able to identify the ongoing inhabitant activity of daily living (ADL) [3]. It is exactly why a growing community of scientists [2,4,5,6] is investigating this specific problem. The issue of recognizing ADL consists in interpreting output signals from distributed sensors and to match them, using a knowledge base, with actions and plans corresponding to the possible ongoing activity. The goal is to circumscribe a minimal set of plausible plans (hypothesis) from the knowledge base by using constraints to eliminate the incoherent hypotheses. Most recognition approaches focus on exploiting only logical [2] or temporal constraints [4] while ignoring the fundamental spatial aspects related to objects in a smart home. Nevertheless, these aspects can play a significant role in the recognition process [7]. In this paper, we propose the integration of topological qualitative spatial relations [9] in an activity recognition algorithm to help reducing