Synthesis of Variable Dancing Styles Based on A Compact Spatiotemporal Representation of Dance

Dance as a complex expressive form of motion is able to convey emotion, meaning and social idiosyncrasies that opens channels for non-verbal communication, and promotes rich cross-modal interactions with music and the environment. As such, realistic dancing characters may incorporate crossmodal information and variability of the dance forms through compact representations that may describe the movement structure in terms of its spatial and temporal organization. In this paper, we propose a novel method for synthesizing beatsynchronous dancing motions based on a compact topological model of dance styles, previously captured with a motion capture system. The model was based on the Topological Gesture Analysis (TGA) which conveys a discrete three-dimensional point-cloud representation of the dance, by describing the spatiotemporal variability of its gestural trajectories into uniform spherical distributions, according to classes of the musical meter. The methodology for synthesizing the modeled dance traces back the topological representations, constrained with definable metrical and spatial parameters, into complete dance instances whose variability is controlled by stochastic processes that considers both TGA distributions and the kinematic constraints of the body morphology. In order to assess the relevance and flexibility of each parameter into feasibly reproducing the style of the captured dance, we correlated both captured and synthesized trajectories of samba dancing sequences in relation to the level of compression of the used model, and report on a subjective evaluation over a set of six tests. The achieved results validated our approach, suggesting that a periodic dancing style, and its musical synchrony, can be feasibly reproduced from a suitably parametrized discrete spatiotemporal representation of the gestural motion trajectories, with a notable degree of compression.