GPS Trajectories Based Personalized Safe Geofence for Elders with Dementia

Abstract

A geofence is a virtual boundary delineated around an area of interest that can be created with a variety of different techniques, such as Wi-Fi, cellular mobile, RFID and GPS. In order to develop a solution that is able to prevent elderly persons with dementia from experiencing elopement or boundary transgression that is frequently related to getting lost or other adverse events if without timely assistive services, in this paper, we propose a data mining based approach to construct a personalized safe geofence by mining individuals' historical GPS trajectories. Specifically, we first model each individuals' outdoor regular movement as a graph based on the fact that most of elders follow a simple life style in routine by visiting relatively fixed places. With this graph model, we can model each elder's outdoor safe geofence as an arbitrary polygon that is defined by the outmost vertexes of the graph model. We then propose a data mining method to construct personalized safe geofence by mining an individual's historical GPS trajectories based on the constructed geofence model, which is separated into three main stages: partitioning GPS trajectories into line segments, extracting characteristic points from a set of the line segments, and instantiating the safe geofence model with extracted characteristic points. A dataset consisting of 3 individuals' GPS trajectory data has been used to evaluate our proposed method. The qualitative results have shown that our method is workable to construct personalized safe geofence by mining individuals' GPS trajectories. Our proposed method has great potential to be used for preventing elders with dementia from wandering to out of the safe area.