What GPS doesn´t tell you: determining one´s context with low-level sensors

As mobile devices become more frequently used by an ever-increasing user base, more efficient methods of use are desired. We are designing a system to use information about the user´s context, including physical, social, and temporal facts, to achieve this, possibly through better user interfaces, application selection, or communication network selection. This paper discusses the structure of the system, including the hierarchy and classification of information. We then describe the working prototype we have built to acquire physical context cues through low-level sensors. Using data collected with this prototype, we show that differentiating between a number of modes is possible using either a standard back-propagating or a modular neural network