The ability of determining and dealing with the trajectories followed by an object in a given (concrete or abstract) space turns out to be quite useful in a variety of contexts. This is the case, in particular, in positioning, where it can be exploited, for instance, for traffic control and user profiling. A key step in trajectory management is the evaluation of trajectory similarity. In many positioning applications, trajectories are built from Global Navigation Satellite System (GNSS) readings; however, in various scenarios, these coordinates are not available. In this paper, we focus on fingerprint positioning systems characterised by a low sampling frequency and a high heterogeneity of the observations. We start with a comprehensive analysis of well-known GNSS-based trajectory similarity measures, and show how some of them can actually be adapted to the fingerprinting setting. Then, we outline a novel approach that exploits multiple information, including both spatial and cellular identifiers with received signal strength. Finally, we make an extensive, experimental comparative evaluation of the various measures (adapted and novel ones) over a real-world fingerprint dataset.
A new similarity measure for low-sampling cellular fingerprint trajectories
Gallo P.;Montanari A.;Saccomanno N.
Primo
2020-01-01
Abstract
The ability of determining and dealing with the trajectories followed by an object in a given (concrete or abstract) space turns out to be quite useful in a variety of contexts. This is the case, in particular, in positioning, where it can be exploited, for instance, for traffic control and user profiling. A key step in trajectory management is the evaluation of trajectory similarity. In many positioning applications, trajectories are built from Global Navigation Satellite System (GNSS) readings; however, in various scenarios, these coordinates are not available. In this paper, we focus on fingerprint positioning systems characterised by a low sampling frequency and a high heterogeneity of the observations. We start with a comprehensive analysis of well-known GNSS-based trajectory similarity measures, and show how some of them can actually be adapted to the fingerprinting setting. Then, we outline a novel approach that exploits multiple information, including both spatial and cellular identifiers with received signal strength. Finally, we make an extensive, experimental comparative evaluation of the various measures (adapted and novel ones) over a real-world fingerprint dataset.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.