Sensor Networks progressively assumed the critical role of bridges between the real world and information systems, through always more consolidated and efficient sensor technologies that enable advanced heterogeneous sensor grids. Sensor data is commonly used by advanced systems and intelligent applications in order to archive complex goals. Processes that build high-level knowledge from sensor data are commonly considered as the key core concept. This paper proposes a semantic layer that would optimally support the knowledge building in sensor systems as well as it enables semantic interaction model at different levels (module, subsystem, system). The semantic layer proposed in the paper is currently used by several architectures and applications in the context of different domains.