Hidden Markov Model 을 이용한 스마트폰 다중 동작에서의 보행 항법 이동방향 개선 알고리즘

Abstract

In this paper, we propose a solution for mismatching between the walking direction and device heading in a PDR (Pedestrian Dead Reckoning) system using a smartphone and the HMM (Hidden Markov Model). The PA Parametric Approach)-based PDR system consists of step detection, step length estimation, and heading estimation. In general, the device heading is regarded as the walking direction, but device heading and walking direction are sometimes mismatched due to the smartphone user assuming various poses. In these cases, the PA-based PDR produces an position error from the mismatched angle. The solution proposed in this paper is an IA(Integration Approach)-based PDR system and a walking direction estimation method using HMM. The IA-based PDR system estimates the position by integrating the acceleration twice. To correct the error produced by this system, the travel distance and the walking direction are used as measurements, and are obtained from the step length model of PA and walking direction from HMM, respectively. Through experiments, it is shown that the performance of the proposed algorithm can overcome the limitations of existing smartphone-based algorithms.