Efficiency of applying virtual reference tag to neural network based RFID indoor positioning method

With the growth of context-aware applications, indoor location positioning receives much attention. RFID is one of the most widely adopted wireless positioning technologies. This paper applies Radial Basis Function Neural Network (RBFNN) to estimate locations of objects based on RFID signal strengths. The architecture of RBFNN is selected via a minimization of the Localized Generalization Error (L-GEM) which selects RBFNN with better generalization ability to unseen samples. Virtual reference tags are adopted to improve positioning performance without additional cost and RF interference. A nonlinear interpolation algorithm is proposed to calculate Received Signal Strength (RSS) for Virtual reference tags which is more reasonable than linear interpolation algorithm. Simulation experiments show that the proposed method outperforms existing method based RFID and virtual reference tags based on nonlinear interpolation algorithm improve positioning accuracy.