In this paper we present a novel probabilistic approach to activity recognition. Our approach is to estimate posterior probabilities of different activities using Bayes' rule. The approach can handle any type of activities as long as it is possible to estimate the conditional probabilities of potential observations, and easily scales to large numbers of activities. We test our approach empirically in an environment where observations are GPS signals of users moving around in a city.
|Title of host publication||2014 11th Workshop on Positioning, Navigation and Communication, WPNC 2014|
|Publication status||Published - 2014|