An effective location fingerprint model for wireless indoor localization

A model for predicting precision and accuracy performance of indoor fingerprint based positioning systems is very desirable for system designers as it helps estimate the probability of location selection before actual deployment. Such information can be used to tune the fingerprint database or improve the offline fingerprint collection phase. This paper presents a new analytical model that applies proximity graphs for approximating the probability distribution of error distance given a location fingerprint database using WLANs received signals, and its associated statistics. Simulations are used to validate the analytical model, which is found to produce results close to that from simulations. The model permits an analysis of the internal structure of location fingerprints. We employ the analysis of the fingerprint structure to identify and eliminate inefficient location fingerprints stored in the fingerprint database. Knowledge of where the inefficient fingerprints are can potentially be employed in a better location fingerprint collecting scheme from a grid system in the offline phase.