Abstract

TempIO is a method to classify a device as either being inside or outside based on its ambient temperature. It takes advantage of the fact that inside temperatures are normally controlled to within a range comfortable for people, while outside temperatures fluctuate with the weather. Inside/outside classification could be used to automatically turn off a GPS receiver when inside, for automatically adding metadata to digital photos, and for higher-level context inference. TempIO works by measuring the ambient temperature and looking up the current outside temperature via a network. We derive a Bayes-based classification rule based on probability distributions of inside, outside, and measured temperatures. Based on test data from five U.S. cities, TempIO classifies correctly 81% of the time when using a web service for outside temperatures and almost 91% of the time when using an ousider thermometer.