Stratified, Computational Interaction Via Machine Learning

R. Murray-Smith

Abstract:We present a control loop framework which enables humans to flexibly adapt their level of engagement in human–computer interaction loops by delegating varying elements of sensing, actuation and control to computational algorithms. We give examples of the use of deep convolutional networks in: modelling and inferring hand pose, single pixel cameras for vision in non visible wavelengths and in a music information retrieval system. In each case we explore how the user can adapt the nature of their closed-loop interaction, depending on context.