A route identification algorithm for assisted living applications fusing WLAN, GPS and image matching data

摘要

This paper addresses the automatic identification of often traversedroutes for assisted living applications using WLANudtechnology in addition to other modalities. This problem isudcomplicated by a number of factors, including the changingudand noisy nature of the WLAN channel, the need to trackudusers seamlessly in both indoor and outdoor environments,udthe need for robustness to slight deviations in the precise path taken, and speed, along a route. In this work commonly traversed routes are identified by clustering based on sensed data, two of which take the form of wireless signals: GPS and WLAN. The latter is particularly important as it can be used both indoors and outdoors. In addition an efficient image matching algorithm is implemented to process data from images automatically taken along the route. In this work a finite number of routes were identified within the DCU campus.udEach route was traversed many times over a period ofud6 weeks and data sequences collected automatically on eachudoccasion. Each such traversal of a route is referred to as a trip in what follows. Section (2) outlines the use of Multidimensional Time Warping in order to automatically cluster trips corresponding to specific routes based on wireless and image data sensed on each trip. Section (3) outlines the manner in which data was sensed and presents clustering results for each modality individually as well as results based on a fusion of the data.