Topic models for RFID data modeling and localization

Topic Modeling, widely used and associated with text-mining applications, is a methodology that may be applied to other domains where the data generation process may be treated as a mixture model, and the underlying data objects modeled similarly to unigrams in the text-mining context. In this paper, we consider the application of topic-modeling to the RFID data generation process, and show how the resulting topics from topic-modeling are applicable to the RFID localization problem - particularly in challenging environments that have significant signal reflections or multi-path.