With the advent of the era of pervasive computing, human dependence on sensors is growing. Such sensor-aware contexts usually possess the characteristics of large-magnitude and uncertainty, so networked devices or systems driven by these contexts cannot get distinct and precise related context information yet, and then lead unable to provide users with satisfied desirable services. In order to solve above-mentioned problems, a two-phased method is proposed in this paper: (1) Context fuzzy processing: Transform a probability-based context into formal context, use fuzzy equivalence matrix clustering method to make the equivalence of the formal background divided, and then get fuzzy formal concepts; (2) Fuzzy formal concept rough processing: Use rough set theory to deal with fuzzy formal concepts, then get implicit-useful context information. Preprocessing through this two-phased method can not only enrich context information and but also make them more comprehensive and complete.