Abstract
While indoor navigation systems attract more and more attention, smartphone-based dead reckoning systems heavily depend on the user holding the device in aspecific way. As soon as a user puts the device in her trouser pocket or jacket, they fail. In this thesis, we enable smartphones to accurately determine steps and the current bearing of a user even in these conditions.
We build a general model of the walking motion with specific instances for different device locations (trousers pocket, jacket pocket). Slicing the measured data within a step to only extract the segment which provides the most information on the user’s walking direction, our approach has a median absolute error of only 12° (mean: 22°, q75: 25°) on a total of 15 participants completing a total of 49 test runs. Therefore, together with proposed feedback generation methods, this thesis lies the foundation of true hands-free indoor navigation without the need for any infrastructure.
Authors
Jens Helge Reelfs