Improved Smartphone-based PDR Localization for Arbitrary Placement

Abstract

Accurate heading information is one of the most important parameters for pedestrian dead reckoning (PDR) techniques. However, existing smartphone-based PDR methods need pedestrian to keep the smartphone in horizontal posture to detect the pedestrian heading, which is not practical. In this paper, we propose a placement-independent pedestrian heading estimation method based on the gradienter. Horizontal posture is common for the placement of mobile phone, and we keep checking the placement. Once the mobile phone is detected in the horizontal posture, the accurate heading information is obtained. According to the turning detection module, the real-time pedestrian heading is matched by integrating the environment information and access point (AP) deployment information. Experimental results show that the error per 100m can be controlled within 2.7m while the smartphone is in arbitrary placement. This method shows good practical value in fast fingerprint database construction.