SoundMark: Accurate Indoor Localization via Peer-Assisted Dead Reckoning

Abstract

Pedestrian dead reckoning enables pervasive indoor localization without a site survey on fingerprints or an intensive deployment of infrastructures. But accumulated errors in dead reckoning limit the spread of pervasive indoor location-based services. Existing landmark-based approaches mostly rely on resetting the user’s position with the landmark position only when the user is detected while on arrival at a landmark. However, such methods are still restricted by the specific movement patterns and sparse landmark distributions so that the opportunity for position calibration is limited. In this paper, an accurate peer-assisted localization system (called <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">SoundMark</i> ) on a smartphone with no prior infrastructure or fingerprinting is proposed. It calibrates mobile user’s dead reckoning position by leveraging the location constraints between another stationary user who arrives at a landmark. To detect whether a user arrives at a landmark, motion pattern is extracted by fusing the multiple sensors. Then, user activity in the landmark is decomposed to determine whether the user is stationary for performing audio ranging. Besides, SoundMark also applies a mobility-induced time-difference-of-arrival-based audio ranging to extract the location constraints between the peers for localization. SoundMark is implemented on the Android platform for evaluations. The results show that the accuracy of proposed peer-assisted localization is within 2.1 m at the percentage of 80%.