Embodied Design of Dance Visualisations

This paper presents the design and implementation of a software platform for creating interactive visualisations that respond to the free-form movements of a non-professional dancer. The visualisations can be trained to respond to the idiosyncratic movements of an individual dancer. This adaptive process is controlled by Interactive Machine Learning. Our approach is novel because the behaviour of the interactive visualisations is trained by a dancer dancing, rather than a computer scientist explicitly programming rules. In this way IML enables an `embodied' form of design, where a dancer can design an interactive system by moving, rather than by analysing movement. This embodied design process taps into and supports our natural and embodied human understanding of movement.

We hope the process of designing an interactive experience for free form dance will help us to understand more about how to create embodied interfaces and allow us to build a general frame- work for embodied interaction. We would also like to create a compelling, embodied and enjoyable experience with more satisfying interactions than previous dance computer games which use pre-scripted routines where a player must repeat a sequence of moves.

The system was developed using a participatory methodology, with a software developer and an interaction designer working in partnership with users to test and refine two prototypes of the system. A third prototype has been built but not yet tested.