Novel Robust Indoor Device-Free Moving-Object Localization and Tracking Using Machine Learning With Kalman Filter and Smoother

Abstract

A novel robust device-free moving-object localization and tracking approach incorporating a machine-learning classifier with both Kalman filter and Kalman smoother is proposed in this work. The received signal-strength indicator (RSSI) maps for an arbitrary indoor geometry are simulated by the prevalent Feko channel-propagation emulator. Such instantaneous RSSIs of WiFi signals collected by different receivers indoors are acquired as a feature vector. The indoor geometry is first partitioned into a number of equisized zones, and then the underlying localization problem can be converted to the multiclassification problem. The RSSI feature vectors are collected as the training data for the gradient boosting decision-tree classifier. We propose a novel online tracking and offline localization approach using the advanced machine-learning technique regularized by Kalman filter and Kalman smoother, respectively. Simulation results demonstrate that our proposed new localization and tracking approach leads to the highest localization-accuracy of 73.1% over four test trajectories as the room is partitioned into 16 equisized zones. Moreover, our proposed new method also reaches the smallest average Euclidean-distance error of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$1.33{\kern0.99594pt}\text{m}$</tex-math></inline-formula> . Our proposed novel indoor localization and tracking scheme would be very promising and convenient as it can well serve for devices free indoor surveillance in the future.