Integrated Statistical Test of Signal Distributions and Access Point Contributions for Wi-Fi Indoor Localization
Mu Zhou,Yaohua Li,Muhammad Junaid Tahir,Xiaolong Geng,Yong Wang,Wei He
DOI: https://doi.org/10.1109/tvt.2021.3076269
IF: 6.8
2021-05-01
IEEE Transactions on Vehicular Technology
Abstract:With the broad deployment of Wi-Fi networks, the Received Signal Strength (RSS) based Wi-Fi indoor localization has attained much interest of both academia and industry. At present, most of the currently available Wi-Fi indoor localization techniques focus on increasing the localization accuracy. However, few of them take into account the diversity of Wi-Fi signal distributions and the measurement error associated with RSS values owing to the complicated indoor environment, which consequently results in the low robustness of indoor localization systems. Thus, with the motivation to tackle this gripping problem, we design a new hybrid hypothesis test based on the idea of Asymptotic Relative Efficiency (ARE), which exploits signal distributions by considering different Access Point (AP) contributions to the Wi-Fi indoor localization accuracy. In concrete terms, first of all, the Jarque-Bera (JB) test is used to perform the normality test on the Wi-Fi signal distribution at each Reference Point (RP), and then the Chi-squared Automatic Interaction Detection (CHAID) approach is applied to obtain each AP contribution degree. Secondly, based on the evaluation of the JB test on the Wi-Fi signal distribution, the hybrid Mann-Whitney U and T test is applied to find the set of matching RPs corresponding to each newly-collected RSS data. Finally, the target location estimate is acquired by using the K-Nearest Neighbor (KNN), where the contribution degree of each AP is assigned as the weight during the calculation to find matching RPs. From the extensive experimental results, it is evident that the proposed approach can successfully improve the system performance by achieving a higher localization accuracy and enhanced robustness when compared with the state-of-the-art Wi-Fi indoor localization techniques.