Indoor localization through dynamic time warping

Identifying and locating oneself in different hallways of high rise buildings forms the classic indoor localization problem. GPS does not work indoors and WiFi may not be omnipresent. This paper presents a novel approach to ambient magnetic fields based indoor localization. We present a system that classifies magnetic signatures using dynamic time warping. Specifically, by aligning similar magnetic signatures that differ in magnitude or time, we classify the signatures and infer the location irrepective of the person and his/her mode of commuting. A Nexus One smartphone was employed, utilizing its builtin magnetic field sensor to create a user friendly localization application solely on the phone. By using a variety of subjects including sighted, blindfolded and people using wheelchairs to handle the human speed variation problem, we evaluated the system across 26 and 15 hallways of two different buildings and obtained accuracies of 92.6%, and 91.1% respectively. With these encouraging results, we believe our proposed solution is user independent and caters to a wide range of hallways.