Super-Resolution Based Fingerprint Augment for Indoor WiFi Localization

WiFi fingerprinting-based indoor localization system is extensively researched with the advent of the high-density wireless networks deployment, but is limited by heavy site survey in the offline phase, for which fingerprint augment is an effective solution. In this paper, we innovatively propose a fingerprint augment method based on super-resolution (FASR) and formulate the processing framework. In order to perform super-resolution on sparse fingerprint database, the conversions between WiFi fingerprint data and fingerprint images are proposed. EDSR, a method based on deep learning in super-resolution, is adopted in FASR to obtain high-resolution fingerprint images, which are then reconstructed to augmented fingerprint database. Experiments on simulated and real scenarios verified the feasibility of FASR. Our work demonstrates a new application of machine learning in wireless communication.