window (Hanning) with weighing coefficients 1, 3,
4, 3, and 1 was used for smoothing raw threedimensional
marker trajectories before computing
the joint angle motion. The gait cycles were extended
or compressed in time to yield a normalized
gait cycle of 64 equally spaced data points. All gait
cycles were expressed as a function of a unit (100%)
cycle length irrespective of the actual time for a
stride. Three out of four cycles of data from each
test session were selected and the mean and standard
deviation for each joint angle pattern were
computed for each subject. Since the subjects were
evaluated on three different days, a total of nine
data sets for a particular subject were averaged,
yielding a representative pattern of motion data for
that individual. Both right and left limb data were
grouped separately. Further, the mean and standard
deviations at each point of the gait cycle were determined
by averaging the mean joint angle data of
all of the subjects.