A novel approach towards utilizing Dempster Shafer fusion theory to enhance WiFi positioning system accuracy

We present a novel approach wherein, firstly, RSSI distribution using kernel density estimation method is approximated via Support Vector Regression (SVR). Secondly, a method is presented to obtain the fused probability mass corresponding to different access points using Dempster Shafer (DS) theory. The fused probability mass thus obtained is used to identify the highly probable points or location of a user over pre-calibrated points. Experimental results demonstrated that the proposed two-step sequential approach is able to limit the 90th percentile localization error to be within meter level. Further the localization accuracy improved to about 88% using our SVR approach in comparison to Gaussian assumption.