Papers

Identifiable Decomposition of Submovements in Human Hand Trajectories

The authors present Sub-ID, a submovement decomposition method for human hand trajectories that uses a spatiotemporal kernel correlation as an identifiability criterion, and it is described as a preprint. On synthetic tests with 3, 8, and 15 overlapping primitives it recovers ground-truth parameter distributions where baselines degrade or fail, holding position error to 0.012 m at K=15 and keeping critical primitive overlaps to about 3%. This gives robotics and motor-control researchers a practical way to extract physiologically grounded motion primitives from long, three-dimensional movements.