Examples from six fundamental dance movement categories, showing reconstructed joint positions over time alongside root trajectories and normalized vertical displacement.
This project explores how machine learning can recognize fundamental dance movements from reconstructed three-dimensional joint trajectories. We compare direct movement recognition, movement-category classification, and a hierarchical approach that identifies the category before the specific movement, evaluating how well each approach generalizes to unseen dancers.
Yanelly Mego
Chapman University
Orange, California, USA
Concepción Valdez
CICESE
Ensenada, Baja California, Mexico
Franceli L. Cibrian
Chapman University
Orange, California, USA
Cibrian, F. L., Valdez, C., Mego, Y., Camarillo-Abad, H. M., Kumar, A., & Beltran, J. A. Temporal Signatures of Expertise: Characterizing Dancers from Non-Dancers Using Wearable Data. Available at SSRN 6026018.