Service discovery is an essential technique to enable applications to find available resources in environments. Since different environments have different resources and distinct environmental characteristics, it is tough to dynamically make use of appropriate services among a number of substitutable candidates according to different contexts. Recently, research efforts on service discovery have worked on abstract representations, semantic matching, and context-awareness. Representing abstractly the capability of services maximizes the availability of the services. Additionally, context-awareness enables the user``s requirements and intents to be richly reflected in service discovery processes, so it makes user distractions minimized. However, abstract representations inherently adhere to a high level of semantics, while the real semantics are dynamically changed according to different contexts. That is, some instances acquired in real environments might not be permitted or not exercise great influence on target users in certain situations. Consequently, there have existed some problems that can arise when integrating abstract representation, context-awareness, and semantic matching into service discovery. In this paper, we propose an activity policy-based service discovery framework for pervasive computing. Human activities are a good starting point to enable computing environments to understand the user’s requirements and intents in human-centered aspects. The correlation between one activity and another activity provides some deterministically critical clues for the proposed scheme to select appropriate services in pervasive computing environments. To take advantage of the beneficial information effectively, in this paper, we introduce $\It{activity policy}$ which includes different resource constraints according to locations, humans, and activities. Through various activity policies, abstract services in developer``s point of view can be dynamically bound t...