Fully Wearable , Stand-alone Person Tracking for Long-term At-Home Monitoring Demonstration of the ActionSLAM System

This document describes the demonstration setup for ActionSLAM, a fully wearable person tracking system that applies action recognitions to correct for the open-loop error accumulation of pedestrian dead reckoning. In various realworld experiments, the algorithm was shown to robustly track people during their daily-life activities. ActionSLAM works in single and multi-floor homes, and it is computationally simple enough to run on two cores of a Samsung Galaxy SIV. Demo participants will try out our Android implementation of the algorithm and build local maps of the environment during a short walk with a motion sensor attached to the foot. The final maps and the user’s reconstructed track can be displayed on the phone user interface. In addition, the application provides audio feedback when the person revisits places during his walk. Keywords—Wearable, Indoor Localization, SLAM, Action Recognition, At-Home Monitoring