Accurate automatic recognition of activities in various contexts is a topic of great interest in the research area of ubiquitous computing. Data-based and knowledge-driven approaches, or a mixture of both, have been developed in the recent years to achieve activity recognition in smart offices, homes of elderly and cognitively impaired and other kinds of buildings. A major problem in this field is the modelling effort, required to adapt a model to a new scenario with new activities, making the practical deployment of such systems often challenging. We discuss how we can apply a planning technique, that is able to represent the hierarchical and structural relationships between activities, namely Hierarchical Task Network (HTN) planning, to some of the state of the art activity recognition models and even improve the gaps in their expressive power. To prove the capability of HTN planning for activity recognition, we propose a generic and simple HTN planning domain model, achieving comparable results with some of the more complex data-driven approaches, which gives a solid reason for further research.