Validation of Predictive Dynamics Tasks - Human Motion Simulation: Predictive Dynamics - page 228

0% height 20% height 40% height 60% height 80% height 100% height

FIGURE 9.21
Predicted model showing human posture during the selected key frames.

FIGURE 9.22
Experimental model showing human posture during the selected key frames.

function. Or, a major problem at a joint will make the modeler add, delete, or
relax some constraints at these joints.
As shown in Figure 9.15 , some participants gave low scores to the experimen-
tally driven avatar. One reason for this poor evaluation is related to the difficulties
associated with the level of visual perceptions and deceptions of humans when
they watch moving objects in the virtual world. Another reason for this rating
could be the way the avatar looks. As shown in Figures 9.21 and 9.22 , the avatars
used for the animation have very detailed realistic skin and human-like features.
These features may distract the observers and cause them to unintentionally check
other irrelevant avatar attributes. For example, the participants were asked to
focus on and evaluate the lifting determinants, but some participants commented
on the way the avatar's head was moving. In general, the first benchmark test pro-
vided significant information about the acceptance of the model; however, it pro-
vides limited information regarding the accuracy of the determinants. This is
because the model has a large DOF with strong coupling between the determi-
nants, which makes it very hard for normal eyes to differentiate between the pre-
dicted and experimental determinants.