A robust multi-cue blending-based approach for floor detection

The task of indoor localization has been carried out with different approaches, which utilize data integrated from distinct sensors belonging to mobile devices. Floor determination is one of the crucial challenges encountered during indoor localization. The solutions to floor determination are mainly based on the techniques of Wireless Fingerprint and RFID (Radio Frequency Identification) sensors. However, the accuracy of such methods still needs to be improved, especially for complex multi-floor building environments with few APs (Access Points). To enhance the accuracy associated with floor determination, in this work, we propose a robust approach for floor detection based on fusion of the RSSI (Received Signal Strength Indication) values and data acquired from an acceleration sensor. Our approach has two key phases. Firstly, the floor pertaining to the user's initial position is automatically generated by using the RSSI data. Secondly, the estimated floor number is verified by utilizing the data acquired from acceleration sensors. With the aid of sufficient experiments on our challenging datasets, we demonstrate that our proposed approach yields more accurate results than state-of-the-art approaches used for determining the floor number.