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