Parameterizing interpersonal behaviour with Laban movement analysis — A Bayesian approach

Abstract

In this paper we propose a probabilistic model to parameterize human interactive behaviour from human motion. To Support the model taxonomy, we use Laban Movement Analysis (LMA), proposed by Rudolph Laban [11], to characterize human non-verbal communication. In interpersonal communication, body motion carries a lot of meaningful information, useful to analyse group dynamic behaviors in a wide range of social scenarios (e.g. behaviour analysis of human interpersonal activities and surveillance system). Taking the advantage of interpretation of social signals defined by Alex Pentland [19], and the descriptive body movement analysis proposed by Laban, we identified characteristics allowing both works to complement each other. To explore in group dynamics, we attempt to show the existent connections between Pentland's descriptions for Interpersonal Behaviours (IBs), and LMA parameters for human body part motions. Those relations are the keys to characterize the interpersonal communication. Given the uncertainty of the phenomenon, Bayesian's methodology is applied. The results present LMA parameters as reliable indicators for IBs, allowing us to generalize the model.