Indoor positioning of mobile devices by combined Wi-Fi and GPS signals

This paper proposes a new heterogeneous interior positioning system using Wi-Fi and GPS signals. It uses the continuous Naive Bayes classifier to incorporate different sources of signal in a single probability estimate. The algorithm assumes that the signals are distributed normally in each location, and the parameters of these distributions (i.e. mean value and standard deviation) depend on the location only. Its accuracy of room-level positioning reached 97.1%, which exceeds the performance of other systems (Redpin, WASP, ProbIN) tested under similar conditions. This system does not require any auxiliary infrastructure, but due to its heterogeneous nature can benefit from it. For example, stationary sensors with locations known a priori can be used to provide a constantly updating reference for the signal levels in different locations. The system can also easily incorporate other sources of signals such as cellular signal or stationary Bluetooth beacons.