| Place awareness is a fundamental component of a pervasive, context-aware system. A common approach to indoor localization is to use signal strengths from nearby 802.11b access points to estimate a position and then determine place based on position. These systems rely on a pre-built signal strength map relating position to a signal strength vector. This thesis proposes a system that is aware of place but does not require an a priori signal strength map relating a position (coordinates) to a signal strength vector. Instead, the system determines place by comparing observed signal strength vectors to a Bayesian model trained on data previously collected at known locations. This paper describes novel algorithms used to estimate place and presents experimental results. In addition, it discusses how the described techniques can be used to create a personal place awareness system that is simple enough to be deployed and maintained by a single end-user. |
