Accuracy improvement of ESPRIT-extracted directionof- arrival estimates using least mean-squares filter for passive RFID inventory application

This paper presents the use of least mean-squares filter to enhance the accuracy of the extraction of direction-of-arrival (DoA) estimates of a passive radio frequency identification (RFID) tag through Estimation of Signal Parameters via Rotational Invariance Techniques (ESPRIT) with a tag-reader system model that makes use of a two-element uniform linear array (ULA) in an inventory system. The use of the adaptive filter least mean squares algorithm (LMS), with step-sizes of 0.01, and 0.005, is designed to reduce the effect of backscattering noise and carrier leakage before extracting the DoA estimates by using the ESPRIT algorithm. The establishment of initialization in the simulation model done in MATLAB® involves random complex signals from angle measures negative 90 through positive 90 degrees measured on the broadside of the ULA. The addition of additive white Gaussian noise modeled as a random variable is included as the backscattered signal as it is received by the antenna array. The comparison of the results of the extracted estimates are done in reference to the actual DoA of the tag by measuring the difference in degrees as root-mean-square error. Simulated observations have also been executed in the case when signal-to-noise ratio (SNR) of the received signal is varied, and when the number of iterations of the filter is increased to present the speed of convergence of the estimates to the desired angle reading of the passive RFID tag.