Spectral Analysis for Dance Movement Query and Interpolation

Emily Napier, Gavin Gray, Sageev Oore

Abstract: While it is possible to quantify motion using various transforms or computational techniques, these representations may not be easy to interpret or reconstruct. In this paper, we focus on the problem of visualizing, querying, and manipulating dance components in the form of spectral features. Our first contribution is measuring the similarity of movements in a way that is robust to phase differences while identifying motions with similar pose frequencies. Our second contribution uses the similarity of pose spectra as a metric to drive the interpolation of a motion sequence towards target statistics. We identify the visual impact of these metrics on the characteristics of motion with input from experts in the dance field. These techniques are implemented to explore representations of dance that have the potential to be the basis of more intuitive choreography generation and educational tools for dance artists.