iPil: improving passive indoor localisation via link-based CSI features

Passive indoor localisation acts as a key enabler for various emerging applications such as secured region monitoring, smart homes, intelligent nursing, etc. Despite of years of research, their accuracy of localisation still remains unsatisfactory for practical uses. The main hurdle lies in the coarse measurement of wireless channels, e.g., received signal strength indicator (RSSI), employed in most existing schemes. In this work, we explore the potential of using channel state information (CSI) for fine-grained passive indoor localisation on a single communication link. To achieve high accuracy, we propose a solution based CSI fingerprint and devise two novel localisation estimator approaches suited to different conditions: weighted Bayesian (W Bayes) and the maximum similarity metric (MSM). Compared with RSSI, CSI has demonstrated itself with a high accuracy of location distinction. Experimental results show that our schemes can achieve a higher accuracy.