Sequence-Type Fingerprinting for Indoor Localization

Location fingerprinting is a widely used technique for indoor localization. Traditional fingerprinting methods are generally in point-type, where the received signal strength indicators (RSSI) at discrete points are trained in the offline phase and used as reference in the online phase. This type of radio map requires many human efforts to calibrate and does not take into account the spatial relationship among the discrete points. In this paper, we propose a novel sequence-type fingerprinting method, which is easy to train and provides more accurate and more robust indoor localization performance. The sequential RSSIs are offline trained when a user moves along indoor routes. Then a segmentation method is used to store the radio map as a graph. In the online phase, the user’s sequential fingerprints within a short time window are measured, and then matched with the sequential fingerprints extracted from the graph by subsequence dynamic time warping algorithm. We conduct simulation to verify the performance improvement of sequencetype fingerprinting over traditional fingerprinting methods.