Abstract
Experiences in pervasive computing environments often rely on estimates of the location and/or behaviour of participants to tailor interaction and/or content to the particular user context. In this paper, we present a novel computer vision method for the indirect monitoring of user position and behaviour. The technique classifies activities based on measurements of the disruptions those activities cause in the surrounding ambient light field. The potential advantages of the method are increased privacy, an ability to recognise high level behaviours and events, and applicability in dimly lit environments. Initial experiments show the technique to be capable of recognising user position within an enclosed environment with at least 75% accuracy. A study of time-based events further shows it to be capable of discriminating between different paths taken through an enclosed space by different sized groups of users with 84% accuracy. We conclude by discussing possible applications of the technique in pervasive computing environments.