Indoor Region Localization With Asynchronous Sensing Data: A Bayesian Probabilistic Model

Obtaining accurate location information of tracked objects is the cornerstone of providing high-quality location-based services (LBSs). Recently, passive localization has become an increasingly important research theme, attracting the attention from both academic and industrial communities. However, existing Wi-Fi fingerprinting-based passive localization methods mainly focus on the indoor localization problem and require Wi-Fi detection equipment to scan wireless communications simultaneously. In this paper, we propose the concept of indoor region localization, which aims at estimating the region where an object stayed for a given time interval. Moreover, a Bayesian probabilistic model is presented to cope with this issue. The key novelty of our method lies in that it allows Wi-Fi detection equipment to behave in an asynchronous fashion, which eliminates synchronization concerns when implementing the positioning system. Specifically, it first introduces a time window for each timestamp, and then aggregates all the sensing records in the given time window to obtain an RSSI measurement vector. Afterward, a kernel function is utilized to measure the conditional probability with respect to every fingerprint, and all these probabilities are combined to obtain the posterior probabilities corresponding to different regions. Extensive experiments are conducted on both public and proprietary data sets and our method significantly outperforms all the comparing algorithms in terms of each evaluation metric.