A Novel Floor Estimation Method in Cellular Networks Based on PCA and Adaboost

Indoor location-based service is getting more and more attention. However, most of indoor localization methods paid attention to the horizontal localization in one floor and little concerned the floor estimation in one building vertically. Generally, base stations around a building can be seen as being in a horizontal plane. Due to near or far to these base stations, position changes in the same floor will introduce different reference signal receiving power(RSRP) change separately. But position changes in the vertical direction will make RSRP change together, which makes RSRP difference not obvious to achieve good floor estimation. Therefore, in this paper we propose a novel floor estimation method in the cellular networks within one building. In offline phase, based on principal component analysis (PCA) and Adaboost, the collected cellular signals are used to train a floor estimation model, where we propose a floor division strategy to improve the floor estimation accuracy. In online phase, the cellular signals are utilized by the floor estimation model to get the floor estimation results. The experiment results show that the floor estimation accuracy of the proposed method outperforms others.