Abstract
In this paper, we present a novel approach for generating a variety of complex behaviours. Our approach is model-based. We train a dual Hidden Markov Model (HMM) on 3D motion capture (MoCap) data representing a number of interactions between two people. Then we track 3D motion of a person in ordinary 2D video. Finally, using the dual HMM and the Viterbi algorithm, we generate a moving "virtual friend" reacting to the given motion of tracked person.