Improving classification performance of RFID gates using Hidden Markov models

Many applications ranging from warehouse logistics to DVD rental require the robust localization and tracking of goods on the item level. RFID represents a popular technology to perform these tasks but suffers from false read events - especially in the widely used UHF range. This paper presents a probabilistic model for an RFID gate. The novelty of our approach is the incorporation of both the uncertainties introduced by the random nature of reader inventories as well as process knowledge into a single well-defined probabilistic framework based on HMMs. The robustness of the resulting tag detection algorithm against stationary tags in the vicinity of the gate is shown experimentally using a demonstrator RFID gate setup.