Pergamon J. Biomechanics, Vol. 28, No. 8, pp.

985 -988, 1995

Copyright 0 1995 Elsevier Science Ltd
Printed in Great Britain. All rights reserved
cm-9290/95 $9.50 + .oo

TECHNICAL NOTE

ERRORS IN ALIGNMENT OF CENTER OF PRESSURE AND FOOT

COORDINATES AFFECT PREDICTED LOWER EXTREMITY TORQUES

Steven T. McCaw* and Paul DeVitat

*Department of Health, Physical Education, Recreation and Dance, 5120 Illinois State University,
Normal, IL 61790-5120, U.S.4 TDepartment of Excercise and Sport, East Carolina University,
Greenville, NC 27858, U.S.A.

Abstract-The purpose of the study was to quantify the effect of errors in spatial alignment between the
center of pressure recorded from a force platform and the coordinates of the foot recorded from film on
resultant joint torques in the lower extremity during the stance phase of gait. Two-dimensional kinematic
and kinetic data from eight subjects performing walking and running were analyzed using inverse dynamics
with the obtained center of pressure values and with f 0.5 and k 1.0 cm shifts in the anteroposterior
location of the center of pressure under the support foot. Shifting the center of pressure posteriorly increased
the flexor (dorsiflexor) torques at the hip and ankle and decreased the extensor (plantarflexor) torques at
thesejoints. This shift also caused an increase in the extensor and a decrease in the flexor torques at the knee.
Shifting the center of pressure anteriorly caused the opposite effectsat each joint. The + 0.5 and k 1.0 cm
shifts in the location of the center of pressure caused, on average, 7 and 14% changes, respectively, in
maximum joint torque and angular impulse values. Relative transition times between flexor and extensor
torques were either increased or decreased, on average, by 7 and 13%, respectively for the two conditions
over all trials. Based on these results, it is concluded that due to potential errors in the spatial alignment of
kinetic and kinematic data, joint torques in the literature on gait should be considered as approximations of
the true values.
Keywords: Inverse dynamics; Limb alignments; Center of pressure; Resultant moment; Kinematics;
Kinetics.

INTRODUCTION
An inverse dynamic analysis is commonly used to calculate
the joint torques in the lower extremity during gait. The
analysis combines data describing the anthropometrics of the
performer, the kinematics of the movement, and known
external forces according to the linear and angular analogs of
Newtons Second Law. Potential sources of error when
calculating joint torques include inaccurate estimations of
segmental anthropometrics, misidentification of segment
endpoints, and inaccurate alignment of the coordinate sys-
tems defining the point of application of the external ground
reaction forces (center of pressure) and the segment endpoint
coordinates.
In spite of an increase in computation of the joint torque
for clinical and research use, few studies have investigated the
effects of potential error sources on the calculated joint
torques. Cappozzo et al. (1975) assessedthe error introduced
in the joint torque calculation from several sources, and
stated that the major determinant of torque errors during
walking stemmed from aligning the ground reaction force
vector with the foot. Nissan (1980) reported that the calcu-
lated knee torque was extremely sensitive to the identified
location of the joint center. A 1 cm shift in the anteroposter-
Received in final form 30 October 1994.
Address correspondence to: Dr Steven T. McCaw, Dept of
HPERD, 5120 Illinois State University, Normal, IL 61790-
5120, U.S.A.
ior location of the knee joint center caused a 15% change in
the magnitude of the knee joint torque during the stance
phase of walking.
Although the misalignment between kinetic and kinematic
data has been identified as the primary source of error in
calculatingjoint torques (Cappozzo et al., 1975), the effect has
not been adequately quantified. The center of pressure,
computed from torques and forces recorded by the force
platform, represents the origin of the resultant ground reac-
tion force vector. Error in accurately locating the center of
pressure on the foot arises because the ground reaction force
and kinematic data are collected using independent measure-
ment systems, often at different sampling rates. The kine-
matic and kinetic data must be aligned both spatially and
temporally in the analysis. Operationally, spatial alignment
refers to accurately digitizing the position of the human in
relation to the force platform. Although these sources of error
are known, their effect on the joint torques have not been
adequately quantified.
The purpose of the study was to quantify the effect of errors
in spatial alignment between the center of pressure recorded
from a force platform and the coordinates of the foot
recorded from film on resultant joint torques in the lower
extremity during the stance phase of gait. An understanding
of this error effect is important when comparing the joint
torques from different test conditions or those presented by
different research laboratories which use different alignment
procedures. It must be noted that this question is entirely
separate from the problem created by errors in the identified
location of the center of pressure on the force platform that
arise from small ground reaction force values during initial
and terminal ground contact.

985
986 Technical Note

METHODS view reduces the film image and increases the error associ-
ated with the center of pressure.
Kinematic and kinetic data were collected for eight subjects Maximum extensor and flexor torques were tdentitied for
(mean mass: 74 kg; mean age: 25 yr) during the stance phases each joint. The plantarflexor angular impulse at the ankle
of walking and running. An average of 9.2 trials were and the extensor and flexor angular impulses at the knee and
analyzed for each subject/condition. Detailed data collection hip joints were also calculated. In addition, the time of
procedures have been presented previously (DeVita et al., transition between positive and negative torques was deter-
1991, 1992), and will be briefly described. mined for each curve. Average values over trials were ob-
A force platform interfaced with a computer was used to tained for each subject/condition and these values were then
record the vertical and anteroposterior ground reaction averaged for the final assessment of each condition. The effect
forces under the right foot and the torque around the of center of pressure location error was determined by
mediolateral axis at a sampling frequency of 1000 Hz. Sagit- calculating the ratio of the parameter values observed with-
tat plane film records of each trial were made with a 16 mm out adjusting the center of pressure to the values obtained
camera operating at a nominal frame rate of 100 Hz. A with the four adjustments. These ratios were expressed in
photoelectric timing system was used to monitor walking percent form.
(1.82 ms) and running (3.83 m s) speeds. Anthropometric
measures including body weight and the circumferences of RESULTS
the upper thigh, knee, ankle and metatarsal heads were
obtained from each subject.
The front corner of the force platform and black circular The anterior and posterior shifts in the location of the
marks placed over the fifth metatarsal head, posterior edge of center of pressure caused similar changes in the joint torques
the shoe at mid-upper height, lateral malleolus, lateral fem- in all subjects for both modes of gait (Figs 1 and 2 and Tables
oral condyle, greater trochanter, and shoulder were digitized 1 and 2). There was little difference between the curves during
in each frame during the stance phase. The data were the initial and terminal 5 and 10% of the support phase for
smoothed with an interactive cubic spline routine and inter- walking and running, respectively. The limited difference
polated to 150 frames of data. The film frames showing initial reflects the relatively small force acting on the foot during
and final contact with the force platform were digitized a these phases of support.
second time to obtain accurate initial and final positions of The posterior shifts in the center of pressure lowered the
the foot relative to the force platform. The distances between torque curve at the ankle, causing a reduction in pIantar-
the posterior edge of the force platform and the heel in the
initial frame and the posterior edge of the force platform and
the toe in the final frame were determined. 210
HIP
The ground reaction force data were similarly interpolated
to 150 points so that each body position during the stance 140 I
phase had a corresponding applied ground reaction force.
The ground reaction force and torque around the mediolate-
ral axis of the force platform were used to calculate the center
of pressure, expressed as a point along the anteroposterior
axis of the kinematic reference frame.
Center of pressure values typically contain error when the
vertical ground reaction force values are low during the
initial and final moments of the stance phase in gait. For the 210.
KNEE
present data, the initial and final four or five center of E
pressure points were usually erroneous. Center of pressure z 140
values for these points were recalculated using the known
positions of the heel and toe relative to the force platform. 70.
For each trial, the initial center of pressure point was set -2
equal to the heel position and the remaining points were
equally spaced between this position and the position of the
first accurate center of pressure data point. A similar proced-
ure was used for the final four or five center of pressure
points.
The Hanavan (1964) anthropometric model and the aver-
age segmental mass proportions presented in Winter (1990)
were used to estimate the inertial properties of the lower
extremity segments. An inverse dynamics analysis, com-
bining the anthropometric, kinematic and ground reaction
force data, was used to calculate the torques at the ankle,
knee and hip joints throughout the stance phase. A conven-
tion of positive torques for extensor and negative torques for
-70 + : : : : : :
flexor was adopted (Winter. 1990). 0 20 40 60 ST---' 100
Additional sets of joint torques were subsequently calcu- RELATIVE TIME i % 1
lated with the center of pressure shifted 0.5 and 1.0 cm
anteriorly and posteriorly, respectively. These manipulations Fig. 1. Joint torque curves during the stance phase of walk-
were chosen because the k 1.0 cm shift was used by Nissan ing from a representative subject. A - 1.0 cm shift in the
(1980) and Cappozzo et al. (1975), and the range of values center of pressure caused a reduction in the ankle and hip
probably include the expected range of error in the center of joint torque curves and an increase in the knee joint torque
pressure location. Error in aligning the center of pressure curve. A + l.Ocm shift in the center of pressure had the
with the subject depends partially on the size of the field of opposite effects.The times of occurrence of the peak values
view of the film image. Although the present study analyzed were not affected but the transition times between extensor
the stance phase only, the field of view was approximately and flexor torques were change&t about 13%, on average. The
4 m wide to include the swing and stance phases, which is the curves for the f 0.5 cm manipulations were intermediate to
more common practice in gait research. This relatively large the curves shown in the figure.
 Technical Note 981

250 center of pressure produced the opposite effectsat each joint,

HIP CPtlcm_---
200 CP as seen in Figs 1 and 2. The adjustments in the center of
I
pressure produced changes in the observed variables of 7 and
14%, on average for the f 0.5 and + 1.0 cm shifts, respect-
ively (Tables 1 and 2).
Although the shifts in the center of pressure did not change
the relative times to the peak torques, the relative times to the
transition between flexor and extensor torques were affected
(Figs 1 and 2). Relative transition times were either increased
or decreased, on average, by 7 and 13%, for the small and

I---
250-
='j KNEE large manipulations, respectively, over all trials.
200.

150.
150 / \ DISCUSSION

2oo:
/
\
100
i/ \ Meaningful comparisons between joint torques from dif-
SO
50~ \ ferent experimental conditions depend on knowledge of the
,/
0. reliability and accuracy of the assessed torques. Inverse
/ dynamic analysis, the method commonly used to assessjoint
torques, has several underlying sources of error which reduce
these.properties. Few empirical observations of the effects of
these errors have been published to assist researchers in
identifying meaningful differences in joint torques. This study
provides an assessment of the error in joint torques due to
error in the alignment between the center of pressure and
kinematic data. This information can be used to help deter-
mine minimum differences between joint torques that can be
considered reliable and meaningful.
Shifting the center of pressure posteriorly increased the
-5Ob : : : : : : : :
20 40 60 80 100 length of the moment arm for the vertical ground reaction
RELATIVE TIME I % I force during the early stance phase, thus increasing the
external plantarflexor torque applied at the ankle. In re-
Fig. 2. Joint torque curves during the stance phase of run- sponse to this increased torque, the dorsiflexor torque appli-
ning from a representative subject. Changes similar to those ed by the muscles crossing the ankle was increased. The
in walking (Fig. 1) were observed when the center of pressure posterior shifts in the center of pressure reduced the length of
was shifted + 1.0 cm. The curves for the f 0.5 cm manipula- the moment arm of the vertical ground reaction force when
tions were intermediate to the conditions shown in this the center of pressure was anterior to the ankle in mid and
figure. late stance, thus reducing the dorsiflexor torque applied to
the foot during the latter period of stance. In response to this
reduction, the plantartlexor muscle torque was also reduced.
The anterior shifts in the center of pressure produced the
flexor torque and an increase in dorsiflexor torque. This shift opposite effects on the moment arm of the vertical ground
raised the torque curve at the knee, reducing flexor torques reaction force and so caused opposite changes in the joint
and increasing extensor torques. The torque curves at the hip torques compared to the posterior shifts.
were subsequently lowered, reducing the extensor torques Since the kinematics, inertial torques and the joint reaction
and increasing the flexor torques. The anterior shifts in the forces were not changed, increased muscle torques at the

Table 1. Mean values for joint torque parameters from walking trials

Center of Maximum Maximum Extensor Flexor

pressure extensor flexor angular angular
shift (cm) torque (Nm) torque (Nm) impulse (Nm s) impulse (Nm s)

Hip
+ 1.0 128.0 - 55.1 12.2 - 4.8
+ 0.5 123.4 - 60.2 11.6 - 5.5
0.0 118.1 - 65.7 11.0 - 6.2
- 0.5 113.7 - 71.1 10.5 - 6.9
- 1.0 108.2 - 76.5 9.8 - 7.6
Knee
+ 1.0 43.6 - 75.0 4.4 - 8.0
+ 0.5 48.5 - 70.4 4.8 - 7.5
0.0 53.1 - 65.1 5.2 - 6.9
- 0.5 57.6 - 60.2 5.5 - 6.3
- 1.0 62.4 - 55.4 5.8 - 5.8
Ankle
+ 1.0 186.5 - 6.5 28.1
+ 0.5 180.8 - 8.1 25.5
0.0 174.7 - 9.8 27.0
- 0.5 168.8 - 11.5 26.5
- 1.0 162.2 - 13.1 25.9
988 Technical Note

Table 2. Mean values for joint torque parameters from running trials
l_-_-..
Center of Maximum Maximum Extensor Flexor
pressure extensor flexor angular angular
shift (cm) torque (Nm) torque (Nm) impulse (Nm s) impulse (Nm s)
-~--
Hip
+ 1.0 256.1 - 50.2 12.2 -- 4.3
+ 0.5 241.3 - 56.0 11.3 - 4.9
0.0 228.8 - 61.9 10.2 -- 5.4
- 0.5 214.2 - 67.6 9.3 -- 6.0
- 1.0 200.3 - 13.6 8.3 -- 6.6
Knee
+ 1.0 201.0 - 52.3 21.3 -- 1.0
+ 0.5 222.1 - 48.0 23.9 - 0.9
0.0 245.5 - 43.4 26.3 - 0.8
- 0.5 261.3 - 38.9 28.9 -. 0.7
- 1.0 289.7 - 34.4 31.4 - 0.6
Ankle
-t 1.0 271.3 -
7.0 35.0
+ 0.5 249.1 -
7.8 32.4
0.0 227.2 -
8.7 29.8
- 0.5 206.0 -
9.6 27.2
- 1.0 183.2 - 10.4 24.6

ankle caused decreased torques at the knee followed by
increased torques at the hip. Reduced muscle torques at the
ankle caused increased muscle torques at the knee, and
reduced muscle torques at the hip.
The single largest source of error in the center of pressure
data results from spatial alignment of the foot and force
platform coordinates (Cappozzo et al., 1975). In the present
procedure, the force platform was digitized in each frame
along with the subject. Additionally, the orientation of the
foot on the force platform was further estimated by a second
round of digitizing which specifically located the initial and
final points of contact between these bodies. Few studies (e.g.
Simpson and Bates (1990)) have described the procedures
used to spatially synchronize the foot and force platform,
making it difficult to assess the accuracy of the center of
pressure and kinematic data alignment. As presently shown,
errors of f 0.5 and 1.0 cm in the center of pressure produce
relatively large errors (7 and 14%, on average) in resultant
joint torques. It is reasonable to assume that center of
pressure data may be in error within the range presently
investigated due to the mtmerous sources of error in esti-
mating exact body landmarks and force platform locations
through film or video digitizing processes.
Based on the results of the study it is concluded that the
joint torque curves presented in the literature may represent
a general profile of the torque pattern at a joint. A previous
study (DeVita and Stribling, 1991) indicated that when trials
were completely reanalyzed multiple times, maximum values
in joint torque curves varied by up to 15%, a percent
variability similar to that reported when the joint centers
were systematically varied by & 1.0 cm (Nissan, 1980). The
15% value is similar to the variability in this study resulting
from manipulation of the alignment between the center of
pressure and the foot. The torque data available in the
literature should be considered as approximations of the true
values. Due to error in locating the center of pressure and
segment endpoints, it is reasonable to assume that published
torque data describing gait may be in error between 7 and
14% of the observed values.
REFERENCES
Cappouo, A., Leo, T. and Pedotti, A. (1975) A general
computing method for the analysis of human locomotion.
J. Biomechanics 8, 308-320.
DeVita, P., Blankenship Hunter, P. and Skelly, W. A. (1992)
Effects of a functional knee brace on the biomechanics of
running. Med. Sci. Sports Exert. 24,797-N%.
DeVita, P., Hong, D. and Hamill, J. (1991) Effects of asym-
metric load carrying on the biomechanics of walking.
J. Biomechanics 24, 1119- 1129.
DeVita, P. and Stribling, J. (1991) Lower extremity joint
kinetics and energetin during backward running. Med.
Sci. Sports Exert. 23, 602-610.
Hanavan, E. P. (1964) A mathematical model of the human
body. Aerospace Medical Division, Wright-Patterson Air
Force Base, OH.
Nissan, M. (1980) Review of some basic assumptions in knee
biomechanics. J. Biomechanics 13, 375-381.
Simpson, K. J. and Bates, B. T. (1990) The effects of running
speed on lower extremity joint moments generated during
the support phase. Int. J. Sport Biomech. 6, 309-324.
Winter, D. A. (1990) Biomechanics and Motor Control of
Human Movement. Wiley-Interscience, New York.