Motion Capture is a method of animating 3D characters by using the movements of real people. This saves on the time needed to key-frame all the motions of the limbs of the body, which is a time consuming process. Motion Capture, or mocap, is used all over the motion picture industry now and world famous movies such Avatar, Happy Feet and Lord of the Rings have used mocap to produce realistic character animation. The method produces movements which are very life-like but which in turn has resulted in some controversy as to whether animations of this kind are indeed animations! (Read more about this here)

There are two principle methods of motion capture:

Marker-Based Motion Capture

In its original form motion capture used markers which are attached to various parts of the body of an actor. The markers are just little highly reflective balls made from a lightweight foam material. Surrounding the actor are infra-red lights which illuminate the markers. As the actor moves the markers are tracked by several cameras and through a process of triangulation the position of the markers in space is stored as (X,Y,Z) coordinates. So, each marker has a stream of (X,Y,Z) data associated with it (markers are identified in an initial calibration phase) while the actor performs some action.

This image above shows the original actor with markers glowing, the extracted 3D points, the application to a basic biped, and finally the application of the data to an avatar

Eventually, these streams of data can be attached to a biped skeleton that controls the movement of a 3D character in, say, a computer game like Assassins Creed II. It may be that upto 3 such markers are used to record the motion of each bone in the biped. In the video below a mocap actor describes the procedure...

Markerless Motion Capture

Marker-less methods track the actor without the use of infra-red sources and markers. Eventually they will be a cheaper alternative to mocap that will revolutionize the human-machine interaction interface.

The general method for this is still based on video tracking: several cameras are used to record an actor performing some action. Using a process of triangulation again the actors body is matched with a template body of similar size and build. The joints of the actor can then be identified. Streams of data can then be associated with these joints as in marker-based mocap.

Here's are example of how this will be used in a multimedia home environment...