A Statistical Indoor Localization Method for Supporting Location-based Access Control

Location awareness is critical for supporting location-based access control (LBAC). The challenge is how to determine locations accurately and efficiently in indoor environments. Existing solutions based on WLAN signal strength either cannot provide high accuracy, or are too complicated to accommodate to different indoor environments. In this paper, we propose a statistical indoor localization method for supporting location-based access control. First, in an offline training phase, we fit a locally weighted regression and smoothing scatterplots (LOESS) model on the signal strength received at different training locations, and build a radio map that contains the distribution of signal strength. Then, in an online estimation phase, we determine the locations of unknown points using maximum likelihood estimation (MLE) based on the measured signal strength and the stored distribution. In addition, we provide a 95% confidence interval to our estimation using a Bootstrapping module. Compared with other approaches, our method is simpler, more systematic and more accurate. Experimental results show that the estimation error of our method is less than 2m. Hence, it can better support LBAC applications than others.