Journal of Biomechanics 32 (1999) 615}621

Technical Note

A marker-based measurement procedure for unconstrained wrist

and elbow motions
Ralf Schmidt*,1, Catherine Disselhorst-Klug, Jiri Silny, GuK nter Rau
Helmholtz-Institute for Biomedical Engineering at Aachen University of Technology, Pauwelsltrasse 20, 52074 Aachen, Germany
Received in "nal form 18 January 1999

Abstract
A protocol is proposed to obtain the joint angles of wrist and elbow from tracked triads of surface markers on each limb segment.
Cu!s placed on the limb support the rigidity of the triads. Additional markers are used to mark the approximate positions of joints.
Corrections of surface marker data for skin motion are derived from a priori knowledge about plausible joint motions. In addition,
ill-conditioned states are trapped when the elbow is nearly fully extended. The protocol is applied to sample motions which
demonstrate the use and the e!ect of the corrections. The results show that the model assumptions are reasonable and that accurate
joint rotations can be obtained. The correction steps prove to be an essential part of upper-extremity movement analysis. ( 1999
Elsevier Science Ltd. All rights reserved.
Keywords: Motion analysis; Upper-extremity; Joint angle; Kinematics; Skin motion

1. Introduction
Diagnosis and treatment of many orthopedic and
neurological disorders could bene"t from motion analysis of the upper-extremities. However, measurement procedures for upper-extremities lack the quality of available
gait analysis systems due to the higher number of degrees
of freedom (dof ) and the range of motion which causes
large skin movements. Therefore, the measurement and
analysis of upper-extremities is usually simpli"ed by
constraining the movement (Andrews and Youm, 1979;
Happee, 1992) or by neglecting some dof (Whiting et al.,
1988; Atkeson and Hollerbach, 1985).
Starting with the common assumption of a rigid-body
model (Andrews and Youm, 1979; Nigg and Herzog,
1995; Hatze, 1980) a marker-based optical measurement
of the movements is desired. Key problems are the relative motions between markers of one segment and relative motions between markers and bones (Fuller et al.,
1996; Reinschmidt, 1996).

*Corresponding author. Tel.: 0049 241 807235; fax: 0049 241 88

88442; e-mail: rschmidt@hia.rwth-aachen.de.
1 Partially supported by Oxford Metrics, Ltd.

For the analysis of body motions usually Cardan

angles are calculated from the marker trajectories. This
requires the proper de"nition of joint axes (Nigg and
Herzog, 1995; Chze and Dimnet, 1995). The orientation
of these axes and their stability during a movement is
essential for an accurate analysis.
Former approaches to analyze upper-extremities
movements did not allow a wide range of applications
since they were invasive (Hogfors et al., 1988) or the setup
restricted the movements (Andrews and Youm, 1979;
Peterson and Palmerud, 1996). The aim of the proposed
measurement procedure is to yield a non-obstructive
setup for a high-quality assessment of all rotational dof of
free wrist and elbow movements.

2. Method
2.1. Kinematic model and setup
The rigid-body model of the upper-extremities consists
of three segments: the upper-arm, the forearm and the
hand, connected by two ball-and-socket joints: the elbow
and the wrist. Both joints possess three rotational dof
while translations are ignored. Another simpli"cation of

0021-9290/99/$ - see front matter ( 1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 1 - 9 2 9 0 ( 9 9 ) 0 0 0 3 6 - 6

616

R. Schmidt et al. / Journal of Biomechanics 32 (1999) 615}621

the model is that the pro-/supination of the forearm is

performed in the elbow only and not in the two radioulnar joints.
The required minimum of three non-collinear markers
per segment is used to measure all six dof of every
segment (Fig. 1). Inter-marker motions are suppressed by
connecting each three markers of a segment. The forearm-cu! and markers are "xed close to the wrist only in
order to record most of the pro-/supination.
The joint centers and joint axes are de"ned with additional markers at the acromion, at the lateral and medial
epicondyli and lateral and medial to the wrist #exion axis
(Fig. 1).
2.2. Static reference measurement
A static reference measurement is used to measure the
locations of the joint markers and centers with respect to
the segment markers and to de"ne the neutral joint
orientations. After this measurement the joint markers
are removed. This avoids to use marker positions which
are disturbed by large skin movements at the joints.
During the reference measurement the subject holds the
upper-arm vertical, the elbow is #exed and the wrist
straight. The straight wrist will be used to de"ne the
neutral wirst orientation.
The joint center of the wrist at the time t of the static
R
reference measurement is the middle between the ulnar
and radial wrist marker (Fig. 1):
(1)
c (t )"1 (m (t )#m (t )).
3!$ R
W R
2 6-/ R
The elbow joint center is the middle between the medial
and lateral elbow markers:
c (t )"1 (m (t )#m (t )).
(2)
E R
2 .%$ R
-!5 R
The shoulder center is assumed to be 7 cm inferior to the
acromion marker which is the average of visually determined distances using a ruler
c (t )"m (t )!e 7 cm
(3)
S R
!#3 R
z
e is the vertical unity vector of the laboratory coordinate
z
system.
2.3. Dexnition of coordinate systems
The markers de"ne several segment and joint coordinate systems of which the relative motions are evaluated
to calculate the joint angles. The coordinate systems of
the upper-arm, the forearm and the hand are arbitrarily
"xed at each three segment markers, resulting in the
homogeneous transformation matrices T , T
and
UA FA
T . These segment coordinate systems are calculated
HD
from the marker positions with the least-squares algorithm presented by Veldpaus et al. (1988). From the positions of the joint markers at the time t of the reference
R
measurement the joint centers at any time are calculated.

Fig. 1. Marker arrangement and joint coordinate systems. Joint

markers are bright and labeled. Segment markers are dark. Joint
centers are dotted.

The center of the wrist and the positions of the wrist joint
markers are given by
(4)
c (t)"T (t)T~1(t )c (t ),
UA R W R
W
UA
m (t)"T (t)T~1(t )m (t ),
(5)
3!$
UA
UA R 3!$ R
m (t)"T (t)T~1(t )m (t ).
(6)
6-/
UA
UA R 6-/ R
The elbow center is calculated from the average of the
elbow positions relative to upper-arm and forearm to get
a better accuracy
c (t)"1(T (t)T (t )#T (t) T (t ))c (t ).
E
2 UA
UA R
FA
FA R E R
The shoulder center moves with the upper-arm

(7)

c (t)"T (t) T~1(t ) c (t ).

(8)
S
UA
UA R E R
Now, the orientations of the wrist and elbow coordinate
systems at any time can be calculated. The axes of the
wrist coordinate system (Fig. 1) are de"ned by
c (t)!c (t)
W ,
x " E
W Dc (t)!c (t)D
E
W
(c (t)!c (t))](m (t)!m (t))
W
3!$
6-/ ,
z " E
W D(c (t)!c (t))](m (t)!m (t))D
E
W
3!$
6-/
y "z ]x .
W
W
W

(9)
(10)
(11)

R. Schmidt et al. / Journal of Biomechanics 32 (1999) 615}621

These are the column vectors of the rotation matrix of

the wrist coordinate system
R (t)"(x
y
z ).
(12)
W
W W W
The axes of the elbow coordinate system are calculated
from the joint center positions.
c (t)!c (t)
E ,
(13)
x " S
E Dc (t)!c (t)D
S
E
(c (t)!c (t))](c (t)!c (t))
E
S
E
,
(14)
y " W
E D(c (t)!c (t))](c (t)!c (t))D
W
E
S
E
z "x ]y .
(15)
E
E E
These are the column vectors of the elbow rotation
matrix
R (t)"(x y z ).
E,1
E E E

(16)

617

The function &arctan2' is the four quadrant arctangent.

The two solutions di!er by 1803.
The rotation angle C around the z-axis is
C"arctan2(R , R cos H!R sin H) .
(21)
21 11
31
The third angle U is the value of the rotation around the
x-axis
U"arctan2(R sin H#R cos H, R sin H
12
32
13
#R cos H).
(22)
33
The ambiguity of two solutions is solved by taking the
angle triple with the smaller sum of absolute values since
wrist and elbow usually do not show extreme joint
angles.
Finally, the signs of the rotation directions must be
de"ned. Flexion, ulnar abduction and pronation, respectively, internal rotation are counted positively. The signs
of the angles U, H and C are changed accordingly.

2.4. Calculation of joint angles

2.5. Corrections
Since joint translations are neglected only the rotational part R of the homogeneous transformation matrices T are used for the calculation of the joint angles.
The orientation of the wrist joint is calculated from the
relative orientation between the hand and the wrist joint
coordinate system.
R (t)"R~1(t) R (t) R~1(t ) R (t ) R (t ).
(17)
83*45
HD
W
HD R W R D R
Since the orientation of the hand coordinate system is
not aligned to anatomical axes the joint orientation is
de"ned at time t with the matrix R (t ). Typically, the
R
D R
joint orientation is straight at the reference time t , which
R
means R (t ) is the identity matrix.
D R
The relative orientation of wrist and elbow joint coordinate system yields the orientation of the elbow joint.
R
(t)"R~1 (t)R (t).
(18)
%-"08
E,1
W
It is not necessary to de"ne the joint orientation of the
elbow at a certain position since both coordinate systems
are aligned with anatomical axes.
To obtain the Cardan angles the rotation matrices
R (t) and R
(t) are decomposed into #exion/exten83*45
%-"08
sion, abduction and rotation around the longitudinal
axis. With the given joint coordinate systems this is
equivalent to the rotation sequence y-, z- and x-axis.

R
R
R
11
12
13
R(t)" R
R
R
21
22
23
R
R
R
31
32
33
"R(y, H(t))R(z, C(t)) R(x, U(t)).

(19)

The rotation angle H around the y-axis is

H "arctan2(!R , R ) or
1
31 11
H "arctan2(R ,!R ) .
2
31
11

(20)

Skin and soft tissue movements cause considerable

relative motions between the bones and the markers.
Although, the directions of the longitudinal segment axes
can be measured reliably the assessment of rotations
around these axes are seriously disturbed by skin movement artifacts. This is because the skin close to the
proximal joint rotates little around the longitudinal segment axis. Only at the distal end of the segment the skin
follows most of the segment rotation. The cu!s guarantee
the same amount of rotation for every marker, but still
they rotate less than the bones.
At the upper-arm this problem is avoided by the de"nition of the elbow joint coordinate system. The orientation of the elbow coordinate system is decoupled from
the rotation of the upper-arm markers around the upperarm axis. However, this de"nition becomes unreliable
when the elbow approaches a straight position. Therefore, a second elbow rotation matrix is de"ned for #exion
angles H )153. This de"nition "xes the last known
E,1
reliable elbow coordinate system at the upper-arm
markers:
R (t)"R (t)R~1(tH* 3)R (tH* 3).
(23)
UA
E,2
UA
15 E,1
15
In the range between 153 and 73 it is interpolated between
the two rotation matrices by weighted averaging between
the rotation angles.
R (t)"R(y, 0 )R(z, c )R(x, u ),
E,1
1
1
1
R (t)"R(y, 0 )R(z, c )R(x, u ),
2
2
2
E,2
0"k0 #(1!k)0 ,
1
2
c"kc #(1!k)c ,
1
2
u"ku #(1!k)u ,
1
2
using k"(H !73)/(153!73).
E,1

(24)
(25)
(26)
(27)
(28)

618

R. Schmidt et al. / Journal of Biomechanics 32 (1999) 615}621

The angles determine the new elbow rotation matrix

which replaces R (t) in Eq. (18).
E,1
(29)
R (t)"R(y, 0 )R(z, c)R(x, u).
E
For #exion angles below 73 only R is used.
E,2
At the forearm the skin movement is explicitly measured and the missing rotation is added to all pro-/
supination motions. To do this a pro-/supination with
straight wrist joint is recorded. The basic assumption
is that during this motion no rotation around the
longitudinal hand axis in the wrist joint occurs. Further,
it is assumed that the hand indicates the exact amount
of pro-/supination. Then any measured rotation around
the longitudinal hand axis in the wrist joint can be
assigned to the incomplete rotation of the forearm
markers.
This skin movement artifact is corrected by adding
a rotation *u(t) around the x-axis to the motion of the
wrist joint coordinate system. This de"nes a new wrist
joint rotation matrix
R] (t)"R (t)R(x, *u(t))
(30)
W
W
which replaces R (t) in Eqs. (17) and (18). The angle
W
*u(t) is calculated as a portion of the measured pro-/
supination with respect to the reference position.
*u(t)"d(U (t)!U (t )).
(31)
E
E R
The optimal factor d minimizes the amplitude of the
rotation around the longitudinal hand axis in the test
measurement. For all other motions "rst U (t) is calE
culated with the original matrix R (t) and then wrist and
W
elbow orientations are recalculated with the corrected
rotation matrix R] (t) .
W
In the next step, misalignments between the anatomical and the introduced wrist axes caused by inaccurate joint marker placement are corrected. Again, a
test measurement is used to assess the misalignment.
The subject is asked to perform a pure #exion/extension
of the wrist. The ideal result would be a changing
#exion/extension angle H (t) only. By introducing
W
a correction matrix R
the wrist joint coordinate
#033
system is rotated to match the ideal case as well as
possible. Eqs. (17) and (18) are extended accordingly.
The "nal equation for calculating the wrist joint
orientation is
R

2.6. Examples
The procedure was tested with 10 subjects (9 male,
1 female), between 22 and 29 years old. A "ve camera
Vicon 370 motion analysis system with a sample rate of
50 Hz was used to record the movements. In a "rst static
trial the joint axes were de"ned with the additional joint
markers. The subjects had to hold their arm in front of
their chest for 1 s. Then the joint markers were removed
and pro-/supination and the wrist #exion/extension
which are necessary for the corrections of the wrist coordinate system were performed. These single-axis movements were executed in a seated position without an aid.
The arm was held in front of the chest during the 5 s
which were allowed to repeat the motion several times.
The subjects were instructed to execute the movements as
accurately as possible at any speed.
These preparing movements are followed by a tracking
task. The subject was seated in front of a plotter. At the
plotter pen a marker was "xed which was to be followed
by the index "nger. An eight shaped curve was traced by
the pen which was created by superimposing the two
sinus functions x"120 mm sin((4n/10s) t) and y"
180 mm sin((2n/10 s) t). The center of the eight was
approximately half a meter in front of the subject's chest
at table height (see Fig. 2).

3. Results
Fig. 3 shows the Cardan angles of wrist and elbow
during the test pro-/supination without and with correction of the skin movement artifacts. There is a large

(t)"R~1 R] ~1(t) R (t) R~1(t ) R (t ) R R (t ) .

HD R W R #033 D R
HD
#033 W
83*45
(32)

The elbow joint orientation is given by

R
(t)"R~1(t)R] (t)R .
(33)
%-"08
E
W
#033
Since the x-axis of the wrist coordinate system is the most
reliable axis R should introduce a rotation around the
#033
x-axis only. But if necessary any other reorientation is
possible.

Fig. 2. View from above on the set-up for the tracking task.

R. Schmidt et al. / Journal of Biomechanics 32 (1999) 615}621

619

Fig. 3. Correction of skin-movement artifacts at the forearm. (a) Original wrist angles showing rotations around the longitudinal hand axis. (b) Wrist
angles, corrected by adding 35% pro-/supination to the original forearm rotation. (c) Original elbow angles. (d) Corrected elbow angles with increased
pro-/supination.

rotation around the longitudinal hand axis in Fig. 3a.

This rotation is caused by the too small rotation of the
forearm markers around the longitudinal forearm axis.
When 35% (d"0.35) of the original pro-/supination is
added to the forearm movement the longitudinal hand
rotation vanishes (Fig. 3b). The rotations around the
other wrist axes are only minimally a!ected by this correction (Fig. 3b and d). The amount of correction was
adjusted individually. The minimum value of the 10 subjects was d"0.17 and the maximum value was d"0.43
(median d"0.26).
In the next step, the static misalignment between introduced and real #exion axis which causes crosstalk in the
#exion/extension movement (Fig. 4a) is corrected. The

crosstalk is signi"cantly reduced when the wrist coordinate system is rotated 43 around the x-axis (Fig. 3b). The
corrections of the x-axis of all subjects ranged from 0 to
73 (median 3, 53).
The tracking movement (Fig. 5) includes rotations
around the longitudinal upper-arm axis. Therefore, this
motion can show the stability of the elbow coordinate
system in spite of skin movements at the upper-arm. An
incorrect abduction does not occur due to the de"nition
of the elbow coordinate system. The largest motion is the
#exion/extension in the elbow. The pronation stays almost constant at 1503. In the wrist joint only small
motions are performed of which #exion/extension is the
major part.

620

R. Schmidt et al. / Journal of Biomechanics 32 (1999) 615}621

Fig. 4. Alignment of the wrist coordinate system. (a) Wrist angles showing crosstalk during a pure #exion movement. (b) Realignment of the wrist
coordinate system reduces the crosstalk.

Fig. 5. Joint angles of wrist and elbow during an eight-curved pointing motion. The #exion/extension in the elbow is predominant.

4. Discussion
The results show that it is necessary to consider skin
movement when all rotational dof of wrist and elbow
movements are to be measured accurately. Several steps
in the presented procedure e!ectively reduce the in#uence of skin-movement artifacts on the joint angles. The
markers of each segment are connected in order to
achieve a rigid body behaviour of the marker de"ned
segments. Even though the bone movements cannot be
observed directly some plausibility tests can be used to
check and correct axes de"nitions. In this way skin movements at the upper-arm and the forearm are corrected.
The method even makes it possible to measure the
amount of skin movement around the longitudinal segment axes.

The amount of skin movements at the forearm seems

surprisingly high. Even with the cu! "xed close to the
wrist the pro-/supination would be underestimated by
17}43% without correction. Looking at the hand rotations during pro-/supination provides a simple way to
correct this skin movement artifact.
Misalignments of the wrist joint axes are corrected by
rotating the joint coordinate system appropriately. So
far, the size and direction of the correction have to be
found by trial and error. This procedure can only work
correctly if the movement is really a pure #exion/extension. The necessary alignments were rather small which
shows the close correspondence between real and introduced rotation axes.
Due to its de"nition this kind of correction is not
necessary for the elbow joint axes. The use of the wrist

R. Schmidt et al. / Journal of Biomechanics 32 (1999) 615}621

center, the elbow center and acromion yields a robust

estimation of the elbow axes, which is decoupled from the
inaccurate marker rotations around the upper-arm axis.
A close approximation of the upper-limb axis is essential
for this. The use of the acromion marker is not the
optimal solution and can a!ect the accuracy of the elbow
#exion angle.
Joint translations are easily derived from the formulas.
However, they are not presented here since "rst analyses
showed that they are at the most in the range of the
measurement accuracy.

References
Andrews, J.G., Youm, Y., 1979. A biomechanical investigation of wrist
kinematics. Journal of Biomechanics. 12, 83}89.
Chze, L., Dimnet, J., 1995. Modelling human body motions by the
techniques known to robotics. In: Allard, P., Stokes, I.A.F., Blanchi,
J.-P. (Eds.), Three-Dimensional Analysis of Human Movement.
Human Kinetics, Champaign, pp. 177}200.
Atkeson, Ch., G., Hollerbach, J.M., 1985. Kinematic features of unrestrained vertical arm movements. Journal of Neuroscience 5(9),
2318}2330
Fuller, J., Liu, L.J., Murphy, M.C., 1996. A comparison of lowerextremity skeletal kinematics measured using skin- and pin-

621

mounted markers. Proceedings of the 4th International Symposium

on 3-D Analysis of Human Movement, University Joseph Fourier,
Grenoble.
Happee, R., 1992. Goal-directed arm movements: I. Analysis of EMG
records in shoulder and elbow muscles. Journal of Electromyogr
Kinesiology 20, 165}178
Hatze, H., 1980. A mathematical model for the computational determination of parameter values of anthropomorphic segments. Journal of Biomechanics 13, 833}843
HoK gfors, L., Sigholm, G., Herberts, P., 1988. Biomechanical model of
the human shoulder: II. The shoulder rhythm. Journal of Biomechanics 24, 699}709
Nigg, B., Herzog, W. (Eds.), 1995. Biomechanics of the Musculo-Skeletal System. Wiley, Chichester.
Peterson, B., Palmerud, G., 1996. Measurement of upper extremity
orientation by video stereometry system. Medical and Biological
Engineering Computation 34, 149}154
Reinschmidt, C., 1996. Three-dimensional tibiocalcaneal and
tibiofemoral kinematics during human locomotion } measured with
external and bone markers. Ph.D. Thesis, The University of Calgary, Calgary
Veldpaus, F.E., Woltring, H.J., Dortmans, L.J.M.G., 1988. A leastsquares algorithm for the equiform transformation from spatial
marker co-ordinates. Journal of Biomechanics 21, 356}360
Whiting, W.C., Gregor, R.J., Finerman, G.A., 1988. Kinematic analysis
of human upper extremity movements in boxing. American Journal
of Sports Medicine 16(2), 130}136