Confidence interval estimation for fingerprint-based indoor localization

Fingerprint-based localization methods provide high accuracy location estimation, which use machine learning algorithms to recognize the statistical patterns of collected data. In these methods, users’ locations can be estimated based on received signal strength vectors from some transmitters. However, data collection is a labor-intensive phase, and should updated periodically. Many researchers have contributed reducing this cost. The easiest way remove cost fingerprints generated by model-based approaches, in trained algorithm environment changes. Probabilistic-based algorithms, addition user location, estimate region interest called 2σ confidence interval probability presence 95%. Gaussian process regression (GPR) probabilistic method that used achieve goal. conventional GPR (CGPR) cannot accurately when noise-free approaches are training phase. paper, we propose novel GPR-based algorithm, named enhanced (EGPR), improves level estimation compared existing while fixing computational complexity online We also theoretically prove minimum variance unbiased efficient estimators. Experiments under line-of-sight non-line-of-sight conditions demonstrate superiority our proposed over counterparts terms as well applicability real-time systems.