Biomedical Engineering Reference
In-Depth Information
where l is the vector containing the lengths of all the body segments, m is the number
of markers and f is the kinematic function that computes the estimation of a marker
placement according to the angles and the lengths of each body segment. As a result,
joint angles are known together with the length of each body segment. Again the
hypothesis here is that the joint are perfect mechanical joints.
In most of the applications in VR high accuracy is not needed when measuring
joint positions in time. However animating avatarswith such inaccurate data generally
leads to artifacts, such as foot skating, flying avatars or collisions with the ground.
8.3.2 Measuring Joint Angles
In most applications involving motion capture data, measuring joint position is only
a first step. Avatars are driven with joint angles and not with positions. Magnetic or
inertial motion capture systems provide the user with global orientation of sensors
attached to body segments. However, because of inaccuracies, local displacements
of the sensor on the body segments and external perturbations, the data provided by
these systems should be corrected (see [
3
,
22
,
33
] for the specific case of magnetic
sensors).
Let us consider now how to compute joint angles according to local reference
frames defined either thanks to the ISB recommendations [
39
,
40
] or the H-ANIM
norm (see
http://www.h-anim.org
for a description). If we use the Euler-like angles
for a ball-and-socket joint, the problem consists in finding the three angles
θ
x
,
θ
y
and
θ
z
that transform the father reference frame F
i
(such as the one attached to the pelvis)
to the child one F
j
(such as the one attached to the thigh), as depicted in Fig.
8.3
.
Fig. 8.3
Local reference
frames associated with two
adjacent body segments: F
i
the parent segment (such as
the pelvis) and F
j
the child
segment (such as the thigh)
