Improvements in sensor technology and small, handheld wireless communication devices provide new opportunities for smart home applications to support independent living for elder care. However, with addition of new sensing technology in the smart home, intelligent methods are needed that process data collected by the sensors to recognize activities for monitoring the well-being of the home's inhabitants. To address this challenge, we designed a smart home system with a multi-agent middle layer to study case-based reasoning methods and constraint satisfaction for activity recognition. The study includes the development of an ontology encoded in RDF to match features of cases and their constraints against observed events in the home. Initial results show that activity recognition can be done successfully using the proposed methods.