Inverse Dynamics for Analyzing Uninstrumented Video of Human Motion

Sommer Gentry, Massachusetts Institute of Technology

Swing dancing is both a historical art form and an interesting functional example of a decentralized communication and control system. The two partners communicate which movements to do via leaning, pushing, and pulling at their points of connection, usually the hands and arms. A few hundred minutes of generally low-quality video constitute the complete record of dances performed in the first half of the 20th century, and so recovery of the lost force data from that video would be valuable to preservationists. Accurate force-torque profiles of swing dancers doing a basic step, the sugar push, can be generated by inverse dynamics using recorded time histories of joint angles and SD/FAST software to generate equations of motion for the bodies. Inverse dynamical analysis required models of the inertial properties of the six rigid body segments, which were generated using the methods of Yeadon (1990). The method was verified by comparison with force versus time profiles recorded using a Tek-scan pressure sensing system.

Abstract Author(s): Sommer Gentry