Pedestrian Positon Estimation Using Sensor Fusion Method Considering Gait Phase

Abstract

In this paper, we propose a method to mitigate the vulnerability of pedestrian dead reckoning (PDR) system to drifted position error through sensor fusion method. The heading of various error factors in general PDR system is unobservable, so it is difficult to reduce the error without supplementary information. Although there are many studies to reduce this heading error, a method using building information is generally used. However, since the building information has to be known in advance, the proposed method is a method using only IMU attached to both shoes. The proposed algorithm is based on consensus filtering used in single target tracking. Assuming that the position of the pedestrian is a single target, two estimation results can be obtained through the sensors attached to each shoe. However, since two inertial sensors estimate the position of each shoe, it estimates the position of the moving shoe. In order to assume a single target, we consider the gait phase so that both estimates can estimate the same position. Experimental results show that using the proposed algorithm improves the performance of the PDR system.