Probabilistic-KNN: A Novel Algorithm for Passive Indoor-Localization Scenario

Lei Yang,Hao Chen,Qimei Cui,Xuan Fu,Yifan Zhang

DOI: https://doi.org/10.1109/vtcspring.2015.7146033

2015-01-01

Abstract:Deterministic methods in indoor-localization systems based on the received signal strength (RSS) almost utilize the average value of the RSS, such as the k- nearest neighbor (KNN) algorithm. However, the distribution of RSS is not always normal Gaussian in the real complex indoor environment so the average value may not represent the location well. To solve this problem, we present a novel algorithm, named as probabilistic KNN (pKNN) algorithm. The algorithm uses the probability of RSS in the Radio-map as a weighting to calculate the Euclidean distance, and it filters the RSS value whose probability is less than 3%. At the same time, we propose a new application environment called as passive indoor-localization scenario. In this scenario, the access point (AP) collects the RSS when the mobile terminal (MT) is not connecting to the AP. Experiment and results analysis for different k values show that p-KNN algorithm is feasible and effective in passive indoor- localization scenario. Finally, comparing to the KNN algorithm, p-KNN algorithm can achieve a better average location accuracy.