The study introduces a new dataset of "free roaming" and "targeted roaming" scenarios, where participants walk around a university campus or locate specific rooms within a library. The accuracy of user identification using a machine learning pipeline with a Radial Basis Function Network (RBFN) classifier is investigated, yielding accuracies of 87.3% for free roaming and 89.4% for targeted roaming. Results show the impact of trajectory length on accuracies, with best results obtained from specific durations cut from the end of trajectories. Higher order velocity derivatives are also explored to enhance identification accuracy. The study compares results with previous research conducted in laboratory settings, highlighting the feasibility and advantages of non-laboratory settings for user identification studies.
Vers une autre langue
à partir du contenu source
arxiv.org
Questions plus approfondies