Pervasive computing applications aim to provide appropriate services that respond directly to their users and environments, with greatly reduced explicit human guidance. These applications usually adapt to changing failure-prone context information which was acquired from various sources that differ in quality and format. To efficiently acquire, share, correlate, and reason over raw existing context data, they must be modeled in a homogenous fashion. In this paper, we propose a hybrid context modelling approach, which attempt to integrate the advantages of object-oriented model and ontology-based model for their distributed context handling and efficient context reasoning respectively. We have applied this model to the development of a context management middleware which providing an extensible application framework for monitoring and assisting the elderly at home environments.