Human movement classification and analysis are important in the research of health sciences and the arts. Laban movement analysis is an effective method to annotate human movement in dance that describes communication and expression. Technology-supported human movement analysis employs motion sensors, infrared cameras, and other wearable devices to capture critical joints of the human skeleton and facial key points. However, the aforementioned technologies are not mainstream, and the most popular form of motion capture is conventional video recording, usually from a single stationary camera. Such video recordings can be used to evaluate human movement or dance performance. Any methods that can systematically analyze and annotate these raw video footage would be of great importance to this field. Therefore, this research offers an analysis and comparison of AI-based computer vision methods that can annotate the human movement automatically. This study trained and compared four different machine learning algorithms (random forest, K neighbors, neural network, and decision tree) through supervised learning on existing video datasets of dance performances. The developed system was able to automatically produce annotation in the four dimensions (effort, space, shape, body) of Laban movement analysis. The results demonstrate accurately produced annotations in comparison to manually entered ground truth Laban annotation.
Guo, W., Craig, O., Difato, T., Oliverio, J., Santoso, M., Sonke, J. and Barmpoutis, A. (2022). AI-driven Human Motion Classification and Analysis using Laban Movement System. International Conference on Human-Computer Interaction. In International Conference on Human-Computer Interaction (pp. 201-210). Springer, Cham. https://doi.org/10.1007/978-3-031-05890-5_16 (Session Chair in S015: Ergonomic Design, Anthropometry, and Human Modeling)