PDRNet: A Deep-Learning Pedestrian Dead Reckoning Framework
Omri Asraf, Firas Shama, Itzik Klein
Abstract
Pedestrian dead reckoning is a well-known approach for indoor navigation. There, the smartphone’s inertial sensors readings are used to determine the user position by utilizing empirical or bio-mechanical approaches and by direct integration. In this paper, we propose PDRNet, a deep-learning pedestrian dead reckoning framework, for user positioning. It includes a smartphone location recognition classification network followed by a change of heading and distance regression network. Experimental results using a publicly available dataset show that the proposed approach outperforms traditional approaches and other deep learning based ones.
Topics & Concepts
Dead reckoningPedestrianComputer scienceHeading (navigation)Deep learningArtificial intelligenceInertial navigation systemPosition (finance)Computer visionPedestrian detectionInertial measurement unitMachine learningGlobal Positioning SystemInertial frame of referenceEngineeringTelecommunicationsPhysicsFinanceTransport engineeringQuantum mechanicsEconomicsAerospace engineeringIndoor and Outdoor Localization TechnologiesVideo Surveillance and Tracking MethodsGait Recognition and Analysis