Indoor localization for an unknown radio emitter employing graph-based optimization and improved RSSD

Accurately locating an indoor unknown radio emitter (URE) is a challenging target to ensure telecommunication security. The URE positioning method based on received signal strength difference (RSSD) has attracted considerable attention due to the advantage of not being affected by transmit power and frequency. However, the RSSD-based fingerprint technique cannot accurately express the constraint equations between signal characteristics and geographic coordinates because of redundant databases and false matching. In this paper, a novel RSSD-based indoor positioning method using factor graph (FG) for an URE is proposed to improve positioning accuracy and reduce computational complexity. Firstly, the databases are reconstructed by singular value decomposition (SVD) to eliminate redundant factor nodes. Secondly, Pearson correlation coefficient (PCC) is utilized to determine the sub-positioning area. Combing SVD with PCC, the hyperplane equations are reconstructed to build an optimized FG model, called SP-FG. Considering simulation and experiment, the proposed SP-FG algorithm improves the cumulative distribution function (CDF) of average positioning error within 0.5 m by 10% and 14% compared with conventional FG algorithm and K-nearest neighbor (KNN) algorithm, respectively. In addition, this paper discusses the superiority of proposed SP-FG algorithm in positioning accuracy under different reference side lengths, access point (AP) coordinates and numbers.