As the aging population grows, age-related diseases show a cohesive increase. Ambient Assistive Living (AAL) systems are being developed and are continually evolving in various areas. While most researchers focus on robust Activity Recognition techniques, this paper investigates some of the architectural challenges of the AAL systems. This study proposes a new system architecture that brings together the service-oriented architecture (SOA), Semantic Web technologies and other methods to address some of the shortfalls in the predecessor system implementations using off-the-shelf and open source components. A partial system implementation is then presented using the proposed system architecture. The system takes some of the key design aspects such as extensibility, reusability, scalability, and maintainability into consideration that can then be seen as a base to further extend the capability of monitoring, collecting, processing and accurately recognising complex or concurrent activities in its overall aim to support assistive living.