Biomedical Engineering Reference
In-Depth Information
A major challenge for validation of predictive dynamics (PD) tasks is the
uncertainty in the predicted motion because people assume different strategies
when performing the same task. This is attributed to differences in strength and
anthropometry, but also to other physiological and psychological parameters that
are beyond the scope of this topic. The variances in strategies will produce differ-
ent joint angle profiles for various people. As a result, it is expected that valida-
tion processes that are based on only statistical measures may not be effective in
capturing such differences.
The second difficulty is associated with the technology for measuring and cap-
turing motion, often referred to as motion capture. For example, a digital human
modeled as having 55 degrees of freedom (DOF), with each DOF having a joint
profile that must be recorded over a time period, may generate an amount of data
that is difficult to handle. Fortunately, in most cases, the real active motion is
conducted by a limited number of DOF that represent a subset of the total DOF.
The motion of the remaining DOF could be considered as passive, and may not
be critical to the validation process.
Another potential difficulty in the validation of PD tasks is the accuracy of
collecting the experimental data and transforming them from the Cartesian space
to the joint space. In this regard, state-of-the-art motion capture systems with
appropriate calibration and a realistic number of cameras should be used to
acquire the subject's motion data using appropriate marker placement protocols.
Robust inverse-kinematic software should be used in calculating the joint angles
from the positional marker data. There are several inverse kinematics (IK) pro-
grams to do that; however, most of them are commercial packages and are limited
to animating their avatars based on Cartesian information. They may not be useful
for animating the predictive human model under investigation (Santos) because it
has a detailed human-like skeleton.
The last, but not least, major difficulty in the validation of motion is human
perception of virtual reality, which comes into effect when people from different
backgrounds are asked to evaluate the motion of an avatar. Two people observing
the same avatar may have different impressions of the realism of the avatar's
motion; this happens even when the avatar's motion is derived with accurate
motion capture systems from human subjects. The role of human perception of
virtual reality will become an important component of the validation methodology
when it comes to discussing what is acceptable motion and what is not.
In this chapter, a framework to validate the predicted motion of a whole-body
task is introduced. The validation framework is based on benchmark tests to char-
acterize the conditions under which the predicted motions are considered accept-
able. Some of the benchmark tests are based on qualitative comparisons and are
used to construct a general perception about the normality of the motion. The val-
idation method and process should be thought of as a mechanism for providing
feedback as well as substantial insight into the PD task.
Additional benchmark tests are based on quantitative comparisons and provide
critical and detailed information about the quality of the model in general and the
Search WWH ::
Custom Search