A Supervised Learning Detector for UHF RFID System

Machine learning (ML) based communication systems are supposed to hold the potential to improve performance. In this work, a supervised learning method with support vector machine (SVM) was proposed for real-time signal detection in UHF RFID system to optimize the system performance. It can achieve frame synchronization detection and raw IQ time-series FMO-encoded data detection with frequency deviation of ±22%. The algorithm was mainly trained off-line using experimental data, and efficiently implemented in a UHF RFID system based on software-defined radios (SDR) platform. The system was validated on real-time signal of commercial tags. Experimental results showed that the “learned” system had displayed improvement in detection accuracy and stableness compared to the existing system based on correlation algorithm.