
136 Dimensionality Reduction
in Figure 9.2. To appreciate the problem’s difficulty, we first need
to understand how motion sequences are represented.
Figure 9.3: A series of skeletal poses in a motion sequence.
An animation or motion sequence typically consists of a series
of skeletal poses, as illustrated in Figure 9.3. The skeleton is de-
fined by a number of bones connected by joints and the poses are
parameterized by spatial or physical attributes including positions,
angles, and velocities associated with the bones and joints. Rich
and realistic motions necessitate a large number of such attributes.
Thus, the animation traces out a high-dimensional