Method and system of crowd- sourced magnetic fingerprinting with neural network re-training

Abstract

Methods and systems of crowd-sourced magnetic fingerprinting for mobile device indoor navigation with neural network re-training are described. In an example, at multiple locations, magnetic input parameters in accordance with a magnetic infrastructure profile of an indoor area are determined. Further, from a mobile device positioned at a first location, measured magnetic parameters at respective ones of multiple locations are received. A route segment similarity factor for the mobile device based on the measured magnetic parameters of a corresponding route segment accessible from the magnetic fingerprint dataset is computed. If the similarity factor exceeds a threshold, the magnetic measured parameters are added to the magnetic fingerprint dataset. For at least a second location along the route, another set of magnetic measured parameters is received from the mobile device. At an output layer of a neural network, an error matrix is computed and an initial weight matrix is recursively adjusted to diminish the error matrix.