What kind of machine learning techniques are usually used to keep track of the markers?
In motion capturing, one usually puts so-called markers onto a person moving (which can be seen here), and films the person with the markers with several cameras from different angles.
At the end, you have sets of points in 3D and because the cameras are filming, you have about 160 snap shots of positions of points per second.
The markers are passive, so you can't really tell which marker moved where from one frame to another. Because you don't have a continuous tracking, but rather discrete snapshots one question in preprocessing the data of a moving person is: Which marker moved where? Also, you might want to classify which part of the body, or in the case of several persons, which marker belongs to which part of which person.
Furthermore, you might have to impute positions since some of the markers might be covered by the body in some frames (e.g. markers on the back when a person lies on the back).
Are there machine learning algorithms which cater to this kind of classification task (classifiying over a time lapse)?