UHF RFID Tags Zoning Using Harmonic Signals and Machine Learning

This paper describes an effective method for improving the detection of UHF RFID (Ultra High Frequency Radio Frequency Identification) tags in a restricted area. The so-called zoning technique is a recurring problem in practical RFID applications: it consists in detecting within an environment with multiple tags that are exclusively present in the zone of interest. The proposed method is based on the concept of Nth harmonic, a new paradigm that involves utilizing the harmonic signals backscattered by tags. Such a method is coupled with a machine learning technique. Experimental results show the importance of harmonic features for better tags zoning. Using a four-layer CNN classifier, we can achieve 99% prediction accuracy by leaving a keep-out distance of 0.5 m between two zones, using the harmonic RSSI (Received Signal Strength Indicator) sum feature, and 94.7% by using the best feature at $f_{0}$ , which is the RSSI max, achieving around 5 times less prediction errors. Furthermore, combining the harmonic and fundamental features leverage the prediction accuracy to 99.8%.