| Abstract |
| During the past decades, many fingerprint-based indoor positioning systems have been proposed and achieved great progress. However, to fulfill the requirement of a real indoor location-based service, many challenges still remain. One major issue is the device diversity where the device used to train the radio map and the device used to sense radio signal strength (RSS) for position query are difference. Due to device diversity, the sensed RSS would be away from the trained radio map and this leads to poor positioning accuracy. To compensate the device diversity effect, a linear transfer model was adopted in our system. However, a practice matter is how to automatically build the regression model during the operation phase. In this paper, we proposed a landmark-based RSS compensation method. Based on the Wi-Fi landmarks, our system could automatically collect RSS pairs between two devices in order to train the transfer model. Our experimental results demonstrated that the proposed system could stably detect the Wi-Fi landmarks. Also the automatically collected RSS pairs perform well if comparing with the manually labeled pairs. With the proposed method, the v RSS difference caused by device diversity could be compensated and hence the positioning accuracy is boosted. |