Estimating Interruptibility in the Home for Remote Communication Based on Audio-Visual Tracking

This paper presents a method for automatically estimating human interruptibility in the home environment. To make online remote communication smoother, determining if it is appropriate to interrupt the remote partner is critical. As a first step in achieving this goal, several audio-visual features, extracted from data streams captured by a camera and a microphone, are correlated to human assessments of own interruptibility. Based on these features, the level of interruptibility is estimated using the trained Support Vector Regression (SVR) technique. Finally, we discuss the potential of our method based on the results of several experiments.