Dynamic dance warping: Using dynamic time warping to compare dance movement performed under different conditions

Dynamic time warping (DTW) is proposed as a technique to assess the difference between two dance performances in terms of timing and to provide further insight into dancer cognition. The DTW method is validated for use with dance performance motion tracking data by comparing its results with 'ground truth' results obtained from a comparison between videos of two motion tracked performances. The technique was extended to investigate two hypothesised processes that affect movement timing-scaling (a fixed ratio alteration) and lapsing (caused by insertion or deletion of movement material). As an example of the use of the technique, an ensemble contemporary dance work was performed with the motion of one of three dancers captured in two conditions - with no music (NM) and with music (WM) - with one repeat of the two conditions. The application of the DTW-based algorithm demonstrates that lapses explained much of the timing mismatch (9.6 out of 14 seconds), with a small proportion explained by scaling (a ratio of 0.976) consistent with previous research. However, after again performing the dance under NM and WM conditions the DTW technique demonstrated a non-trivial contribution of scaling in explaining time differences across the various combinations of conditions. In these comparisons, scaling cannot be eliminated as a possible underlying factor of timing error, and it may be that correct scaling (aiming for a ratio of 1) must be learned via practice.