| 期刊名称： |
IEEE Transactions on Human-Machine Systems |
| 全部作者： |
Baoding Zhou *,Qingquan Li ,Qingzhou Mao ,Wei Tu ,Xing Zhang |
| 出版年份： |
2015 |
| 卷 号： |
45 |
| 期 号： |
5 |
| 页 码： |
562 - 574 |
| 查看全本： |
|
Thispaperpresentsanactivitysequence-basedin-doorpedestrianlocalizationapproachusingsmartphones.Theactivitysequenceconsistsofseveralcontinuousactivitiesduringthewalkingprocess,suchasturningatacorner,takingtheelevator,takingtheescalator,andwalkingstairs.Theseactivitiestakeplacewhenauserwalksatsomespecialpointsinthebuilding,likecorners,elevators,escalators,andstairs.Thespecialpointsformanindoorroadnetwork.Inourapproach,wefirstdetecttheuser’sactivitiesusingthebuilt-insensorsinasmartphone.Thedetectedactivitiesconstitutetheactivitysequence.Meanwhile,theuser’strajectoryisreckonedbyPedestrianDeadReckoning(PDR).Basedonthedetectedactivitysequenceandreckonedtrajectory,werealizepedestrianlocalizationbymatchingthemtotheindoorroadnetworkusingaHiddenMarkovModel.Afterencounteringseveralspecialpoints,thelocationoftheuserwouldconvergeonthetrueone.Weevaluateourproposedapproachusingsmartphonesintwobuildings:anofficebuildingandashoppingmall.Theresultsshowthattheproposedapproachcanrealizeautonomouspedestrianlocalizationevenwithoutknowingtheinitialpointintheenvironments.Themeanofflinelocalizationerrorisabout1.3m.TheresultsalsodemonstratethattheproposedapproachisrobusttoactivitydetectionerrorandPDRestimationerror.
