HiHeading: Smartphone-Based Indoor Map Construction System with High Accuracy Heading Inference

Abstract

Smartphone is widely used in indoor map construction with its build-in sensors. However, the low accuracy problem of build-in sensors always causes the collected user trajectories noisy. The significant problem we face is the low accuracy of walk heading estimated by build-in sensors because of phone heading fluctuation and magnetic field anomaly. This paper presents HiHeading - a high reliable crowdsourcing-based indoor map construction system. HiHeading leverages build-in inertial sensors to construct accurate motion traces. These traces are generated by HiHeading with high accuracy based on the novel ideal of fusing gyroscope and orientation sensor to get reliable walk heading estimation in indoor dead reckoning (DR). To evaluate our system, we have tested it in a middle size indoor office by recording 3 people's walk trajectories during 5 days. We present an evaluation of our system and the experiment result shows 70% of the heading estimation error is lower than 10 degrees.