Wi-fi fingerprinting with reduced signal strength observations from long-time measurements

Abstract

Indoor positioning which uses signal strength values of Wi-Fi networks have become popular as these wireless networks often already exist and many mobile devices, such as smartphones or tablets, have built-in Wi-Fi cards. Usually fingerprinting is employed for positioning which achieves relatively low positioning accuracies on the several meter level. In the scope of this work two methods are presented which have the potential to improve the fingerprinting performance using long-time RSS observations at reference stations. Both methods employ the usage of at least three reference stations surrounding the area of interest on which signal strength observations are continuously performed during the whole measurement process. Thereby the first method uses a 2-D linear plane-interpolation for the deduction of real-time corrections. For that purpose, the measured signal strengths are reduced by the long-time measurements which are interpolated at the approximate position of the measuring point. In the second method the daily average of the long-time measurements is applied and the improvements of the measurements are calculated by the deviation from the daily average. For this method it is conceivable that a single reference station may be sufficient if it is located in the middle of the area of interest. Field tests were performed in an office building and are analyzed. The fingerprinting algorithms reached an averaged positioning accuracy of around 5 m in dependence on the used smartphone. The daily average improvements (DAI) method provided a better performance than the interpolation method which is highly influenced by the required approximate position of the user.