Figure 4

Projection of musical cues (metric classes) onto the dance trajectories. Firstly, (a) the annotation of metric structure of the (b) music is synchronized with the MoCap recording. These cues are projected onto (c) the movement vectors (in the example, right hand movements) as different classes of points (e.g., 1st beat, 2nd beat--respectively described as 1 and 2 in the figure). Finally, (d) the point clouds are discriminated using LDA analysis which guarantees the separation of point-clouds into (e) topologies. In this study we assumed a spherical distribution for the point clouds whose radius is defined by the average of the Euclidean distances from all points of the class to the mean.