Basmag: An Optimized HMM-Based Localization System Using Backward Sequences Matching Algorithm Exploiting Geomagnetic Information

Abstract

Due to the complex building construction materials, the geomagnetic field has great anomalies in indoor environments. These anomalies generate distinctive signatures corresponding to positions. Many hybrid localization schemes that combine geomagnetic information and pedestrian dead reckoning (PDR) have been proposed. In this paper, we regard these distinctive geomagnetic anomalies as fingerprints for high precision indoor positioning. In order to improve the hybrid localization accuracy, we present Basmag, a robust hidden Markov model (HMM)-based indoor localization system, and we also propose a backward sequences matching algorithm (BSMA) to optimize the HMM. In order to improve the low discernibility of single geomagnetic signal, we vectorize the backward consecutive geomagnetic signals to fingerprint sequences with the help of PDR. We match the backward sequences with a pre-constructed fingerprint database to get more precise transition probabilities in HMM. We conducted a theoretical analysis of the feasibility of the BSMA. Extensive simulation results show that Basmag can achieve high-accuracy positioning. Finally, we conducted experiments performed on a smartphone in an indoor area to compare Basmag with other geomagnetic field-based and WiFi-based localization methods, and the experimental results verify the effectiveness and robustness among users with different walking styles of Basmag.