Clinical Biomechanics 14 (1999) 357±360

Accuracy of centre of pressure measurement using a piezoelectric

force platform
James Middleton a,*, Peter Sinclair b, Ray Patton b
a
Moorong Spinal Unit, Royal Rehabilitation Centre, Ryde, Sydney NSW 1680, Australia
b
School of Exercise and Sport Science, The University of Sydney, Sydney NSW, Australia
Received 29 May 1997; accepted 14 August 1998

Abstract
Objective. To evaluate the accuracy in determining centre of pressure when using a force platform for stabilometry research.
Design. Technical report.
Background. Previous studies have determined the accuracy of locating the centre of pressure by applying force through a single
point. In human stabilometry research however, force is distributed across two feet and the centre of pressure located between the
feet.
Methods. Forces were applied to a Kistler force platform using a force applied through a single point as well as when force was
distributed between two metal blocks.
Results. Errors in calculating centre of pressure were substantially larger when force was applied through a single point than
when the force was distributed between two blocks.
Conclusions. Evaluating centre of pressure accuracy by applying force through a single point overestimates the errors inherent in
human stabilometry research.

Relevance
Evaluating changes in the centre of pressure of human subjects standing on a force platform is a common tool for evaluating
control mechanisms of balance. Accurate information regarding the centre of pressure is essential for the validity of this re-
search. Ó 1999 Elsevier Science Ltd. All rights reserved.

Keywords: Force platform; Stabilometry; Balance; Accuracy; Centre of pressure

1. Introduction When using a force platform for stabilometry, force is

Movement of the Centre of Pressure (CoP) with a
subject standing on a force platform has been used to
provide information about postural control in both
normal and pathological situations [1,2].
Bobbert and Schamhardt [3] evaluated the accuracy
of determining CoP by applying forces through a known
point of application to a Kistler force platform. Errors
varied in magnitude up to 20 mm, depending on the
position across the plate. Average errors were 3.5 mm in
the x (short axis) and 6.3 mm in the y (long axis) di-
rection. These errors increased substantially towards the
edge of the force platform, particularly outside the
central region bounded by the four force transducers.
*
Corresponding author.
0268-0033/99/$ ± see front matter Ó 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 2 6 8 - 0 0 3 3 ( 9 8 ) 0 0 0 7 9 - 5
spread over an area of the two feet and the CoP is lo-
cated between the feet. Testing CoP accuracy using a
single point force is convenient, but does not realistically
represent the loading patterns used for stabilometry re-
search. The purpose of this study was to evaluate the
accuracy of CoP calculation using a two `foot' model
that more closely resembled the clinical situation of an
individual standing quietly on a force platform.
2. Methods
A Kistler force platform, type 9281A (Kistler In-
struments AG, Winterthur, Switzerland) was used to
measure ground reaction forces from which CoP data
was calculated using the two di€erent models of force
application. The Kistler force platform was embedded in
358 J. Middleton et al. / Clinical Biomechanics 14 (1999) 357±360

and replaced between each trial. In several positions

along the x and y axes, six trials were collected at each
location. Clear indentations were made by the apparatus
from which Vernier callipers (precision 0.05 mm) were
used after each data collection to measure displacement
of the point of application. Data was collected for 10 s
periods whilst the apparatus applied a static point force
and mean coordinates for the point of application were
calculated from CoP formulas. Coordinates calculated
from the force data (CoPforce ) were then compared with
the measured point of application coordinates
(CoPmeasured ) to assess the accuracy of CoP calculation.

2.2. Stationary `feet' model

Two rectangular model `feet' (illustrated in Fig. 1),

Fig. 1. Model `feet' used to simulate standing and test the accuracy of
CoP measurement under static loading conditions.
a concrete foundation according to the manufacturer's
speci®cations and measured the orthogonal components
of force along the x, y and z axes as shown in Fig. 1.
Raw signals from each transducer were recorded by
personal computer at a sampling frequency of 50 sam-
ples per second via a 12 bit analogue-to-digital converter
(Data Transition DT 2801A).
The point of application of the resultant of the
ground reaction forces was calculated from the orthog-
onal force components using equations from the Kistler
instruction manual.
2.1. Point of force application model
An apparatus was constructed to apply a vertical
force of approximately 300 N through a single point of
application. The pin-point tip of the apparatus was
carefully positioned on the force platform at predeter-
mined positions. Three trials were collected around each
position shown in Table 1, with the apparatus removed
made of steel to the same dimensions (25 cm long and
10 cm wide) and ®lled with lead to approximately
300 N, were used to simulate standing and test the ac-
curacy of CoP measurement.
The weight and CoP of each model `foot' were de-
termined independently by positioning each `foot' as
precisely as possible in the geometric centre of the force
platform. Six separate trials were collected for each
model `foot' and mean coordinates of the CoP and the
weight were calculated individually for each `foot'.
With two model `feet' positioned symmetrically about
the geometric centre of the force platform, the CoP
would not be expected to coincide with the force plat-
form centre owing to the inevitable di€erence in weight
and CoP for each model `foot'. Predicted CoP between
the two `feet' was calculated using the following equa-
tion:
Wtfoot1
CoPfoot1 ‡ Wtfoot2
CoPfoot2
CoPpredicted ˆ
Wtfoot1 ‡ Wtfoot2
(where Wt represents the weight of each `foot' in New-
tons and CoP coordinates are in millimetres).
Trials were collected with the medial edges of the
`feet' symmetrically positioned parallel to the x-axis in
positions illustrated by Fig. 1. These positions were

Table 1
Mean and standard deviation of the di€erences between CoPmeasured and CoPforce in the x and y directions using the single point of force application
model

Point of application (y-axis) Point of application (x-axis)

0 mm 60 mm 120 mm 150 mm

xerror yerror xerror yerror xerror yerror xerror yerror

0 mm 0.6 (0.6) 2.5 (1.2) 1.7 (0.6) 2.9 (1.7) 4.4 (1.0) 2.8 (2.6) 6.2 (1.0) 5.1 (2.8)
50 mm 0.4 (0.3) 3.3 (1.8) 1.9 (0.8) 4.0 (1.7)
100 mm 0.9 (0.4) 3.1 (1.8) 0.9 (0.6) 3.8 (2.3) 3.9 (1.2) 4.8 (2.7)
150 mm 0.5 (0.2) 2.3 (1.1) 1.3 (0.7) 2.2 (1.6)
200 mm 0.9 (0.6) 3.9 (1.6) 3.1 (1.0) 4.2 (1.7)
250 mm 0.9 (0.7) 11.6 (1.1)
 J. Middleton et al. / Clinical Biomechanics 14 (1999) 357±360
chosen to represent typical foot positions during stab-
ilometry research. Alignment was achieved using ®ne
felt tip markings parallel to the x-axis at the desired
displacements. Six trials were performed at each dis-
placement to test reliability. Mean CoPforce coordinates
between the `feet' for each individual trial were com-
pared to the coordinates of the CoPpredicted for each
displacement. Comparison between means in di€erent
positions across the force platform was undertaken to
determine whether any systematic error existed.
Student's t-test was used to compare mean absolute
CoP errors between forces applied by the model `feet'
with the mean for errors applied through a single point
of force. Only single point trials within the region
bounded by the force transducers were used for this
comparison for reasons that will be discussed later.
3. Results
Measurement of the weight and CoP of each model
`foot' was repeatable within 3 N and 1 mm.
Force platform CoP accuracy for the single point of
force application was greatest along the mid-lines of the
force platform and deteriorated as forces were applied
further from the centre. Table 1 shows the absolute
di€erence between coordinates measured directly with
Vernier callipers and calculated from force platform
data.
Accuracy of CoP prediction using the single point of
force model was better for x than y coordinates, with the
magnitude of the mean absolute di€erence (or error) for
all data points collected within the central region
(n ˆ 144) being nearly twice as great in the y-direction
(3.4 mm) as in the x-direction (1.8 mm). Accuracy de-
teriorated towards the edge of the force platform, as
demonstrated by Table 1.
Mean di€erences between CoPpredicted and CoPforce
from trials with the model `feet' are shown in Table 2.
Di€erences never exceeded 1.25 mm and 2.00 mm for
xCoP and yCoP respectively, regardless of `foot' position.
In contrast to the single point of application, there was
no trend for errors to increase as the `feet' moved apart.
Table 2
Absolute value of the di€erence between CoPpredicted and CoPforce be-
tween the two `feet'
Distance between model `feet' xerror yerror
(mm) Mean (SD) Mean (SD)
10 0.4 (0.3) 1.0
100 0.4 (0.2) 1.0
200 1.0 (0.3) 0.5
300 0.7 (0.2) 1.0
400 1.1 (0.7) 0.5
359
Mean error for the two `foot' model was signi®cantly
less than that for the single point of force application
(p < 0:001), even though only single point force results
from the more accurate central region were included in
the comparison. If measurements from the whole force
platform were included, these di€erences would be even
more apparent.
4. Discussion
Single point of force results were found to be consis-
tent with the ®ndings previously described by Bobbert
and Schamhardt [3] with accuracy doubled in the x-axis
compared with the y-axis direction. The distance be-
tween the force transducer positions on the force plat-
forms used in both experiments were proportionately
greater in the y-axis direction by a ratio of approximately
2:1. Calculation of moments therefore results in multi-
plication of errors in force measurements by the same
factor. Consistent with this result was the ®nding that
mean absolute errors in this study were approximately
half those demonstrated by Bobbert and Schamhardt [3],
where a force platform of almost double the dimensions
was used. The forces applied by Bobbert and Scham-
hardt [3] were larger than those for the present study,
however their results demonstrate that larger forces do
not result in greater magnitudes of error.
Also consistent with Bobbert and Schamhardt [3] was
the ®nding that the accuracy of CoP deteriorated mark-
edly towards the edge of the force platform when using a
single point of force application. As their force platforms
were larger, this e€ect was greater for their study.
Despite the apparent potential for error when man-
ually placing the `feet', this model demonstrated much
better accuracy than when using a single point of force
application. Manually re-positioning the `feet' between
trials with alignment of the `feet' using lines drawn with
a ®ne felt-tip pen would be expected to introduce errors
of at least 0.5 mm. Despite this, errors were generally
found to be less than 1 mm, even when the feet were
positioned on the edge of the force platform.
Bobbert and Schamhardt [3] hypothesized that errors
in determination of the point of application using CoP
calculation could be explained by force plate bending
and the fact that the transducers were therefore not
loaded exactly at their centres. Point application of force
loading would be expected to result in greater force
platform deformation than standing with two feet and
hence may overestimate the signi®cance of potential
error during stabilometry tasks. Furthermore, as the
(0.6) `feet' were positioned symmetrically about the centre of
(0.8) the platform, errors resulting from each `foot' deforming
(0.6) the plate would be likely to cancel out. Further inves-
(0.7)
(0.2)
tigation is warranted to determine the e€ect of placing
the `feet' asymmetrically on the platform.
360 J. Middleton et al. / Clinical Biomechanics 14 (1999) 357±360
The present study suggests that the accuracy expected
during two legged stabilometry tasks would be greater
than that previously indicated by single point of appli-
cation loading techniques. Errors in determining the
location of the centre of pressure are likely to be less
than 2 mm when the feet are positioned symmetrically
about the centre of the force platform.
Acknowledgements
This study was supported by a Northern Sydney Area
Health Service grant (No. 94/21).
References
[1] Hufschmidt A, Dichgans J, Mauritz KH, Hufschmidt M. Some
methods and parameters of body sway quanti®cation and their
neurological applications. Arch Psychiat Nervenkr 1980;228:135±
50.
[2] Ekdahl C, Jarnlo GB, Andesson SI. Standing balance in healthy
subjects. Scand J Rehab Med 1989; 21:187±95.
[3] Bobbert MF, Schamhardt HC. Accuracy of determining the point
of force application with piezoelectric force plates. J Biomech
1990;23:705±10.