In this paper, Gleicher talks motion capture: the reasons to use it, various input methods, and why motion editing can be both necessary and hard to do right. He distinguishes between real-time and off-line animation goals. He also discusses alternatives to motion capture, including hand-animation and procedural simulation.

Key ideas: Far from being a simple replacement for trained animators, motion capture is hard to get right, and suffers from cumbersome input methods, potential signal droppoff/noisiness, and large amounts of unstructured data.

Questions: I recall seeing some fairly convincing simple motion capture from video in Vision class -- unfortunately it didn't seem clear to me how well it accounts for relative depth. Is this only useful for 2D capture? Or can it somehow reconstruct 3D information?