n = 28; female, n = 12) there were no significant
differences in the spatiotemporal parameters between
male and female subjects. The overall mean
and standard deviation of angular excursions for the
subjects along with one standard deviation envelope
are shown in Figs. 3-5. The limb rotation angles
are the average of nine cycles from each of the
40 subjects (total of 360 gait cycles). Zero percent
corresponds to the heel strike and 100% corresponds
to the next heel strike of the same limb. The
percent standard deviations for the flexionextension
motion at the hip, knee, and ankle were
smaller than those for the ab-/adduction or internal
and external rotations. The joint angle data also
were further divided according to sex. Except for
hip ab-/adduction, there were no significant differences
between the male and female groups for any
of the joint angle patterns.
The effect of errors in defining the embedded
axes on the computed angles are shown in Fig. 6
using the knee joint as an example. The knee flexion-
extension angle was relatively unaffected while
the knee varus/valgus and rotation angles were affected
nonuniformly throughout the gait cycle. The
results showed that the errors in knee varus/valgus
and rotation angles varied with increasing knee flexion
angle. The magnitude of the errors in the knee
varus/valgus and rotation angles are shown as a
function of the knee flexion angle for different magnitudes
of error in the definition of embedded axes
in Fig. 7A and 7B, respectively. Similar results
were obtained at the hip and ankle joints.
DISCUSSION
In this paper, we have presented a detailed description
and implementation of a technique for
computing lower limb rotations during level walking
using a simple marker system. For computing the
limb rotation angles, a system of axes was defined
based on a set of markers affixed to key anatomical
locations. Two factors were considered in choosing
the anatomical location. The first was to minimize
relative motion between the skin and underlying
bony structures, thereby satisfying the rigid body
assumption. For the skin-mounted markers as well
as the cuff-mounted markers, the rigid body assumption
was found to hold (on the average) to
within 2 3 mm. This did not have a significant effect
on the measured joint angle patterns. The second
consideration was to minimize the amount of manual
intervention needed to sort and track the marker
trajectories accurately. In video motion analysis
systems, it is common for the trajectories of closely
spaced markers to cross each other, thereby making
automatic tracking by the computer extremely difficult.
Manual intervention is often necessary to
identify trajectories of closely spaced markers
whose paths intersect. In gait analysis, the trajectories
of markers placed on the foot present problems
due to their relative proximity to each other.
Therefore, in the present system, only two markers
were used on the foot to define limiting the measurement
of ankle joint motion to flexion-extension
and internakxternal rotation. Due to the geometry
and the size of the foot segment, adding another
marker to measure eversion-inversion angle would
complicate the data analysis. Further, given the finite
accuracy and resolution of the motion analysis
system, the estimates of inversion-eversion may
not be sufficiently accurate to be of any practical
use. By limiting the number of markers on the foot
to two, the time required for data analysis is substantially
reduced, which renders the system attractive
for use in routine clinical gait evaluation.
In any type of motion analysis system, contacting
or noncontacting, a source of error in the estimation
of joint angle motion is due to uncertainty in the
construction of an embedded coordinate system. In
a goniometric system, the alignment of the goniometer
determines the orientation of the embedded
axis. In the present system, the body surface markers
define the embedded axes and therefore their
placement is crucial. While the effect of errors in
the definition of embedded axes on the flexionextension
angles is small, ab-ladduction and rotation
angles are affected significantly. This may be
the reason for the large dispersion reported in the
literature for the knee varus/valgus and rotation angles
and therefore these angles must be interpreted
cautiously. While it may be difficult to define the
embedded axis exactly, it is at least necessary to be