Friday, September 25, 2009

Ambient features beat GPS indoors

Surroundsense [1] implements the simple idea of integrating various ambient features of an indoor location, such as sound, light, and acceleration pattern of the user. The authors present architecture which they test successfully to distinguish between 51 different business locations. In order to deploy that one should check whether assumption that the ambience of places remains stationary really holds. It also remains unclear to me how to actually train such a system...

[1] SurroundSense: Mobile Phone Localization Via Ambience Fingerprinting
Martin Azizyan, Ionut Constandache, Romit Roy Choudhury
ACM MobiCom, September 2009.

No comments: