Biomedical Engineering Reference
In-Depth Information
in the virtual environment and those actually linked to the real movement of the user.
In that case methods based on physical models enable to automatically adapt the
motion to different kinematic and physical constraints [
41
].
However inverse dynamics is sensitive to noise or inaccuracies especially when
computing joint accelerations. Another limitation is that it remains difficult to deal
with closed-loop systems, such as dealing with the double support phase in walking
where the two feet are in contact with the ground. In that case, it is very difficult to
strictly separate the forces exerted below each foot.
8.6.1 Measuring or Estimating Muscle Activities
In some very specific applications, joint torques is not accurate enough to understand
motion strategies. Indeed, joint torques provides us with the resulting action of a
group of muscles whereas control strategies could have a direct link with the action
of one isolated muscle. Slightly changing the axis of rotation of a motion may recruit
different muscle groups even if the resulting joint torque looks the same.
The direct approach tomeasuremuscle activity consists in sensing the electromyo-
grams (EMG) of targeted muscles. EMG is a measurement of the electrical activity
of skeletal muscles recorded with the placement of small electrodes over the skin
(there exist more invasive electrodes but they are unusable for large movements).
Thus EMG is limited to surface muscles. The signal returned by the EMG system is
noisy and required heavy signal processing to estimate muscle tensions in Newtons.
However, it gives an interesting point of view about muscle coordination if several
muscles are measured concurrently. See [
23
] (among many others) for more details
on EMG.
Some researchers have proposed to use indirect methods to retrieve the tension of
all the muscles involved in the studied motion (including deeper muscles). Muscu-
loskeletal models have been introduced in the early nineties [
6
] thanks to the increase
of computation power of computers. The key idea is tomodel muscles thanks to action
lines acting along an axis determined by to muscle insertions. Knowing the accurate
location of each of these muscles and tendons insertion on bones, it is possible to
retrieve these action lines, as shown in Fig.
8.4
.
A muscle is supposed to work only by applying a positive tension (leading to
contractions) and cannot push the bone. Hence for each muscle i, its tension T
i
is
positive. Each muscle is also limited to a maximum voluntary contraction (MVC)
which is generally evaluated in isometric condition (i.e. no displacement of the bones
but exertion of a force against an external load). If we consider the surface of the
cross section area (perpendicular to the muscle fibers), this MVC is given by:
F
max
i
=
K
(
l
)
×
K
0
×
PCSA
where K
0
is a constant ranging from 15 to 33 N/cm
2
according to the authors, and
K(l) is a value that depends on the muscle length l. Indeed there exists a relation
