An indoor localization system by fusing smartphone inertial sensors and bluetooth low energy beacons

Abstract

Indoor localization becomes a research focus in recent years since. Smartphone-based pedestrian dead reckoning (PDR) is one of the widely-adopted localization techniques with limiting problems such as the drift of inertial sensors. Bluetooth Low Energy (BLE) has better performance result which makes it an auxiliary tool for PDR to correct errors. But BLE fingerprint sampling and calibrating are time-consuming and labor-intensive. In this paper, a Support Vector Machine (SVM) classification algorithm based crowdsourcing method is developed and applied to generate BLE landmarks instead of manual work leveraging smartphone sensors and uploaded BLE signals. A particle filter is also used to fusion PDR and landmarks detection results for better localization performance. The experiments show that the proposed fusion algorithm achieved the accuracy of 3.15 m at 90% of the time with dense landmarks (1 landmark per 5 m), which performs 51.76% better than 6.53 m from PDR algorithm. With sparse landmarks (1 landmark per 15 m), the proposed fusion algorithm achieved the accuracy of 3.26 m at 90% of the time. The proposed tracking system using smartphone inertial sensors and BLE beacons can be a promising methodology in practical usages.