Elucidation of the adaptive mechanism of intricate human motion imitated by deep reinforcement learning

2023 Fiscal Year Annual Research Report

• Project/Area Number: 22K20519
• Research Institution: National Institute of Advanced Industrial Science and Technology
• Principal Investigator: Shen Keli 国立研究開発法人産業技術総合研究所, 情報・人間工学領域, 産総研特別研究員 (80965179)
• Project Period (FY): 2022-08-31 – 2024-03-31
• Keywords: TD-PCA / Beat-aligned synergies / Beat detection / Complexity analysis / Deep RL / Motor coordination / Computational modeling / Sports biomechanics

Outline of Annual Research Achievements

We developed a TD-PCA approach to extract beat-aligned motor synergies from street dance datasets, leveraging the first synergy to improve kinematic beat detection and enable accurate beat alignment with music. The enhancement was verified through cross-validation. Our computational framework analyzes complex dance motions, understanding motor mechanisms in dance genres, providing insights into movement dynamics/music relationships, with applications in dance research, performance analysis, training, and injury prevention. We simulated deep reinforcement learning models reproducing dance skills, analyzing them kinematically/kinetically, demonstrating our methods' effectiveness for studying AI-generated dance movements.

Remarks

Keli Shen and Jun-ichiro Hirayama, Beat-aligned motor synergies and kinematic beat detection in street dance movements, eLife (submitted).