Crowdsensing-based organic fingerprint for Wi-Fi localization

Abstract

In a Wi-Fi location system, the accurate fingerprint database is the key to ensure accurate positioning. Several methods have been proposed for database generating and database training, but most of them have been trying to get a static fingerprint database which is only available in stable signal environment. In this paper, we proposed a crowdsensing-based fingerprint updating algorithm which makes the fingerprint organic and guarantee the positioning accuracy both in stable and unstable Wi-Fi signal environments. To get this, we first analyze the Wi-Fi signal environments, the instability of Wi-Fi signal environments will deteriorate the positioning performance of crowdsensing-based Wi-Fi location system if the fingerprint database is fixed. To do fingerprint updating, we first proposed a fingerprint updating algorithm based on the Euclidean distance between distinct fingerprints, which effectively reduce the importing error in crowdsensing-based database generating and signal sampling. The algorithm works well in stable and short-term Wi-Fi environments, but it can't solve the problem when tremendous changes occur in Wi-Fi environments. In order to solve the problem, we proposed an advanced algorithm based on the first one: Fingerprint updating based on Wi-Fi fingerprints' reliability model. In this algorithm, we assign every fingerprint a new property, reliability, which decays with time. By utilizing the reliability of the fingerprint, the location system can automatically detect the change of Wi-Fi environments and remove the outdated fingerprints. With this updating algorithm, we can get an organic fingerprint database.