ARTICLE IN PRESS

Journal of Biomechanics 42 (2009) 23302335

Contents lists available at ScienceDirect

Journal of Biomechanics
journal homepage: www.elsevier.com/locate/jbiomech
www.JBiomech.com

Functional calibration procedure for 3D knee joint angle description using

inertial sensors
J. Favre a,, R. Aissaoui a,b, B.M. Jolles c, J.A. de Guise b, K. Aminian a
a

Ecole Polytechnique Fe de rale de Lausanne (EPFL), Laboratory of Movement Analysis and Measurement (LMAM), Lausanne, Switzerland
Ecole de Technologie Supe rieure, Laboratoire de recherche en imagerie et orthope die (LIO). Montreal, Canada
c
Centre Hospitalier Universitaire Vaudois and University of Lausanne (CHUV), Department of Orthopaedic Surgery and Traumatology, Lausanne, Switzerland
b

a r t i c l e in f o

a b s t r a c t

Article history:
Accepted 15 June 2009

Measurement of three-dimensional (3D) knee joint angle outside a laboratory is of benet in clinical
examination and therapeutic treatment comparison. Although several motion capture devices exist,
there is a need for an ambulatory system that could be used in routine practice. Up-to-date, inertial
measurement units (IMUs) have proven to be suitable for unconstrained measurement of knee joint
differential orientation. Nevertheless, this differential orientation should be converted into three
reliable and clinically interpretable angles. Thus, the aim of this study was to propose a new calibration
procedure adapted for the joint coordinate system (JCS), which required only IMUs data. The
repeatability of the calibration procedure, as well as the errors in the measurement of 3D knee angle
during gait in comparison to a reference system were assessed on eight healthy subjects. The new
procedure relying on active and passive movements reported a high repeatability of the mean values
(offseto11) and angular patterns (SDo0.31 and CMC40.9). In comparison to the reference system, this
functional procedure showed high precision (SDo21 and CC40.75) and moderate accuracy (between
4.01 and 8.11) for the three knee angle. The combination of the inertial-based system with the functional
calibration procedure proposed here resulted in a promising tool for the measurement of 3D knee joint
angle. Moreover, this method could be adapted to measure other complex joint, such as ankle or elbow.
& 2009 Elsevier Ltd. All rights reserved.

Keywords:
Knee
3D Kinematics
Inertial sensors
Functional calibration
Ambulatory measurement
Gyroscopes
Accelerometers

1. Introduction
Three-dimensional (3D) knee joint angle measurement is an
important requirement, notably in the orthopaedic and rehabilitation elds. Although several systems, such as optoelectronic
devices, can make this measurement (Andriacchi and Alexander,
2000; Sutherland, 2002), there is a need for an ambulatory system
that can be used for clinical routine examinations. Since a decade,
some authors have proposed wearable systems including accelerometers, gyroscopes and magnetometers (referred as magnetic
and inertial measurement unit: MIMU) (Foxlin, 1996; Bachmann,
2000; Roetenberg, 2006) or wearable systems composed of
accelerometers and gyroscopes (referred as inertial measurement
unit: IMU) (Luinge and Veltink, 2005; Favre et al., 2006) to
measure the orientation of rigid bodies in space. With MIMUs,
these orientations are calculated relative to an earth-xed frame
dened by the gravity and local magnetic eld. However, MIMUs
are sensitive to the magnetic eld distortions and their use in
clinical setting environments is limited (Roetenberg, 2006;

 Correspondence to: EPFL-STI-IBI2-LMAM, ELH 131/Station 11, CH-1015 Lausanne, Switzerland. Tel.: +41 21 693 76 06; fax: +41 21 693 69 15.
E-mail address: julien.favre@ep.ch (J. Favre).

0021-9290/$ - see front matter & 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.jbiomech.2009.06.025

Picerno et al., 2008). IMUs, in contrary, can be used in

unconstrained environments. Nevertheless, since they have no
sensor sensitive to the heading, the orientation of each IMU is
calculated relative to a particular frame. Actually, this lack of a
common reference frame prevents the monitoring of human
joints. Recently, Favre et al. (2008) proposed a functional
alignment procedure in order to measure the orientation of two
IMUs xed on the thigh and shank segments relative to a common
reference frame. The IMUs of such a system could be small and
lightweight enough to be xed on body segments without
interfering with the execution of movements. However, to be
suitable in clinical applications, the differential orientation
between the thigh and shank IMUs should be described through
three reliable and clinically interpretable angles (Andrews, 1984;
Woltring, 1994; Cappozzo et al., 2005).
For quantifying joint motion, the International Society of
Biomechanics (ISB) recommends the calculation proposed by
Grood and Suntay (1983). Regarding joint angular displacements,
instead of the differential orientation between the IMUs frame,
this model needs as input the differential orientation between the
Cartesian bone-embedded anatomical frames (BAFs). Several
calibration procedures were proposed to transform 3D position
of markers as measured by standard motion capture systems into
BAFs orientation (Cappozzo et al., 1995; Besier et al., 2003;

ARTICLE IN PRESS
J. Favre et al. / Journal of Biomechanics 42 (2009) 23302335

Charlton et al., 2004; Hagemeister et al., 2005). However, since

currently no inertial-based system is able to measure position,
none of these calibration procedures can be used. Recently, some
original calibration procedures were proposed for MIMUs (ODonovan et al., 2007; Picerno et al., 2008). Nevertheless, they cannot
be used with IMUs either.
This study aimed to extend the inertial-based system proposed
in Favre et al. (2008), which allows the ambulatory measurement
of thigh and shank IMUs differential orientation, by proposing a

2331

calibration procedure in order to describe this orientation

according to three clinical angles. Thus, the objectives were: rst,
to develop and assess the repeatability of a new calibration
procedure, which only rely on IMUs data; second, to evaluate the
error by comparing the 3D knee joint angle obtained by the
combination of the inertial-based system and the proposed
calibration procedure to a stationary reference system.
2. Materials and methods
2.1. Description of joint motion

W
The joint coordinate system (JCS) (Grood and Suntay, 1983) was used to
describe the knee joint angles. Fig. 1 shows a joint composed of a proximal (A:
thigh) and distal (b: shank) segment. The BAFs of the segments are respectively IJK
and ijk. The orientation of these two BAFs respective to the xed reference frame
~
~
~
~
(XYZ) are expressed by two quaternions Q
A and Q b . The generalized joint
coordinate e1e2e3 represents the system about which the three clinical rotations
e 2 , and
occur: exion/extension (FE) around ~
e 1 , abduction/adduction (AA) around ~
nally internal/external rotation (IE) around ~
e3.

V
RA

QA

J
I

QA
Z

e3

2.2. Inertial-based measurement system

e1
e2

Qb
k

X
Qb

u
v

b
j

Rb

Fig. 1. The JCS, as well as the axes and orientations denition.

The wearable system was made up of two small IMUs connected to a portable
data-logger (Physilogs, BioAGM, CH). The IMUs were xed to the thigh and shank
segments using an exoskeleton harness (Ganjikia et al., 2000) as shown in Fig. 2a.
The data were recorded at a sampling frequency of 240 Hz. The orientation of the
two IMUs frame UVW and uvw (Fig. 1) were calculated relative to their own static
reference frame using a tracking algorithm described in Favre et al. (2006).
Basically, this algorithm fuses the gyroscope and accelerometer row signals to
calculate an optimised orientation at each time sample. Then, the thigh and shank
IMUs orientation were calculated relative to a common static reference frame XYZ
~0
~0
~
~
and expressed by two quaternions Q
A and Q b (Fig. 1). This alignment was
realised following the functional procedure proposed by Favre et al. (2008), which
estimates the reference frames misalignment comparing the angular velocity
vectors of the thigh and shank segments during a hip AA movement (Fig. 2b). The
combination of the fusion algorithm with the functional alignment procedure was
shown to provide precise short time (up to a few minutes) measurement of thigh

Fig. 2. (a) Measurement systems: the ovals surround the inertial and magnetic sensors. (b) Hip abduction/adduction movement. (c) Passive movement of the shank in its
sagittal plane. (d) Passive movement of the shank in its frontal plane.

ARTICLE IN PRESS
2332

J. Favre et al. / Journal of Biomechanics 42 (2009) 23302335

and shank differential orientation.

2.3. Calibration procedure

The calibration procedure consisted to estimate the constant calibration
~0
~0
~
~
~
~
R b (Fig. 1) which rotate the sensors frame (Q
R A and ~
quaternions ~
A and Q b ) to
~
~
~
~
their corresponding segments BAF (Q A and Q b ):
~0
~
~
~
~
~
Q
A t Q A t  R A

and

~
~0
~
~
~
~
Q
b t Q b t  R b

2.6.1. Repeatability of the calibration quaternions

This repeatability was estimated from the dispersion (w) of the eight
~
~
calibration quaternions (~
R A or ~
R b ) obtained by the movement combinations
(c 1 to N; N 8) around their mean considering all subjects (s 1 to O; O 8)
according to Eq. (2) and (3). This calculation is similar to Della Croce et al. (1999).
v
u
sO
u
cN
u
X
1
u
wt
Dsc 2
2
O  N  1 c1
s1

where t indicates the time and  corresponds to the quaternion multiplication

operation (Kuipers, 2002).
In this study, we considered a rst method (M1) without calibration procedure,
~
~
where we assumed the IMUs perfectly aligned with the segments BAFs (i.e., R
A and
~
~
R b are identity transformations). The functional calibration method (M2) that we
proposed relied on rotation axes and on properties of the JCS through two phases.
First, the shank angular velocity was recorded during two movements performed
by an examiner while the subject was sitting. The rst passive movement
corresponded to a knee FE carried out between 451 and 801 of exion (Fig. 2c). The
second passive movement, was a rotation of the shank in its frontal plane (Fig. 2d)
realized at the hip joint. Based on the angular velocity vectors measured during
these passive movements two mean rotation axes (~
i and ~
j 0 ) were calculated in the
shank IMU frame. In order to obtain a Cartesian ijk BAF, the ~
k axis was dened as
j axis was dened as the normal to
the normal to the f~
i; ~
j 0 g plane and nally the ~
the f~
k; ~
ig plane. Since the ijk frame was dened relative to the IMU frame, it
~
directly corresponded to the ~
R b quaternion. In gait analysis, the lower limbs joint
angles are usually measured during the standing posture in order to dene offsets,
which are later removed from the angles obtained during the walks (Mills et al.,
~
R A was dened in order to have
2007). Thus, in the second phase, the quaternion ~
the three knee joint angles equal to zero during the neutral standing posture.
According to the JCS, the 3D knee joint angle could only be equal to zero when the
~
thigh and shank BAF are collinear. So, the quaternion ~
R A was calculated in order to
align the thigh IMU frame on the shank BAF during the standing posture.

2.4. Reference measurement system

Although several calibration protocols were proposed for the knee joint, none
is currently adopted as a gold standard by the community (Della Croce et al.,
2003; Baker, 2006). Consequently, for this study the calibration procedure
described in Hagemeister et al. (2005) was considered as a reference since,
similarly to the proposed method, it involves functional and postural steps.
However, as the proposed and reference calibration procedures have few
discrepancies (partly due to the measurement systems), some differences in the
3D knee joint angle are expected and will be discussed. The reference system used
two magnetic markers (Libertys, Polhemus, USA) xed on the thigh and shank
parts of the harness (Fig. 2a). During the measurement, an operator followed the
subject, keeping the magnetic source close to the knee joint. The data were
recorded with a sampling frequency of 240 Hz.

2.5. Validation protocol

Eight healthy men (mean age: 26 years old, range 1928) were included in this
study. The harness was xed on their right knee and the IMUs were visually
aligned with the longitudinal axis of the segments. After few minutes to accustom,
the subjects were asked to walk seven meters along a pathway during which the
knee motion was simultaneously monitored by the two systems. The calibration
procedures of the reference and of the inertial-based system were achieved after
the walking periods, since the calibration procedures could modify the harness
position. In order to evaluate the repeatability of the system against the alignment
and calibration movements, the hip AA movement and the two passive calibration
movements were performed twice at 20 min interval. This study was approved by
the Ethics Committee of our hospital and written informed consent was given by
all subjects prior to enrolment.

2.6. Evaluation of the functional calibration procedure

To estimate the repeatability of the proposed system, the six alignment and
calibration movements (twice hip AA in standing posture (mvt1), as well as twice
knee FE (mvt2) and twice rotations of the shank in its frontal plane (mvt3) in
sitting posture) were combined together, leading to eight different movement
combinations (i.e., {mvt11, mvt21, mvt31}, {mvt11, mvt21, mvt32,}, y, {mvt12, mvt22,
mvt32}).

Dsc 2  cos1



~

~ 1 
~
R sc 
R s  ~



!
3
real

where 1 stands for the inverse operation, whereas ||real denotes the real part of the
~
~
quaternion (Kuipers, 2002). ~
R sc corresponds to the calibration quaternion (~
R A or
~
~
~
~
R b ) of the subject s obtained with the movement combination c. R s corresponds,
for each subjects, to the normalized mean calibration quaternion over the eight
movement combinations. DSC represents the absolute angular amplitude of the 3D
~
~
~
rotation aligning ~
R sc with R
s (Spoor and Veldpaus, 1980; Kuipers, 2002).
2.6.2. Repeatability of the 3D knee joint angle
Since the nal purpose of the system is to measure the three knee joint angles,
the propagation of the calibration quaternions repeatability to the knee joint
angles calculation was evaluated. For this, the knee joint angles of the walking
trials (FEsc(t), AAsc(t) and IEsc(t)) were calculated for each subject considering the
~
calibration quaternions (~
R sc ) obtained for each movement combination. Then, the
angles were averaged FE s t; AA s t and IE s t over their eight movement
combinations according to Eq. (4); and similar formulas were considered for AA
and IE:
FE s t

N
1X
FEsc t
N c1

The differences between the eight angles (e.g., FEsc(t)) and their mean (e.g.,
FE s t) were calculated for each subject. The absolute value of the mean differences
(Doffset sc), and the standard deviation of these differences (DSD sc) were then
calculated for every subjects and combinations. Finally, these parameters were
averaged throughout the 8 subjects leading to: mDoffset and mDSD for each knee joint
angle. Moreover, in order to quantify the similarity between the patterns, the
coefcient of multiple correlation (CMC) was calculated between the angles
obtained through the eight combinations for each subject and knee angle (Kadaba
et al., 1989; Kutner et al., 2005). These coefcients were also averaged throughout
the 8 subjects leading to: mDCMC for each knee joint angle. When the patterns are
similar the CMC tends to 1, whereas it tends to 0 if the patterns are totally
dissimilar.
2.7. Evaluation of the 3D knee joint angle
The knee joint angles obtained from the two calibration methods (m M1 or
M2) were compared to the corresponding angles obtained through the reference
system for each subject (e.g., FEs,m(t) vs. FEs,REF(t)). Regarding M2, according to the
results obtained for the Repeatability of the 3D knee joint angle (Section 2.6.2), only
one of the movement combinations (i.e., c 1) was considered. The absolute value
of the mean difference (offsets,m) between the angles (e.g., FEs,m(t) and FEs,REF(t)),
the standard deviation of this difference (SDs,m), as well as the correlation
coefcient (CCs,m) were calculated for each method, subject and angle. Finally,
these three parameters were averaged through the 8 subjects leading to: moffset,m
(accuracy), mSD,m (precision) and mCC,m for each knee joint angle and method.

3. Results
The 3D shank angular velocity measured during the two seated
passive calibration movements are shown for a typical subject in
Fig. 3. The former is expressed in the IMUs frame (which also
corresponds to M1) and in the BAFs which corresponds to M2.
Regarding the functional procedure (M2), the repeatability of
the calibration quaternions (w) was of 2.41 and 2.01 for the thigh
and shank segments, respectively. Table 1 reports the repeatability
of the 3D knee joint angle (mDoffset, mDSD and mDCMC) due to the
propagation of the calibration quaternions precision. These results
showed high repeatability of the patterns for the three knee joint
angles (i.e., small mDSD and high mDCMC). Regarding the inuence of

ARTICLE IN PRESS
J. Favre et al. / Journal of Biomechanics 42 (2009) 23302335

Shank angular velocity, /s

Shank angular velocity, /s

mvt2
200

2333

mvt3
150

sensor frame

sensor frame

100

100

50

0
50

100

100

200

150
0

200

0
150

segment frame

100

100

segment frame

50

0
50

100

100

200

150
0

Time, s

Time, s

Fig. 3. Typical 3D shank angular velocities (1/s) during the two passive calibration movements in sensor and segment frames. mvt2 corresponds to knee FE and mvt3 to
~ or ~
rotation of the shank in its frontal plane. Solid lines corresponds to ~
u or ~
i axes, dashed lines to ~
v or ~
j axes, and dotted lines to w
k axes.

Table 1
Mean (SD) repeatability of the 3D knee joint angle obtained with M2.

FE
AA
IE

mDoffset

mDSD

mDCMC

0.8 (0.9)
0.4 (0.5)
0.6 (0.8)

0.0 (0.0)
0.2 (0.1)
0.1 (0.1)

1.00 (0.01)
0.89 (0.14)
0.91 (0.10)

mDoffset and mDSD are in degree.

calibration movements on the mean angle values (mDoffset), small

differences (less than 11) were noticed for all angles.
Typical 3D knee joint angles obtained with the reference and
the inertial-based system (M1 and M2) are displayed in Fig. 4.
Table 2 presents the errors of the 3D knee joint angle (moffset, mSD and
mCC) obtained through M1 and M2. For the three angles, higher
precisions were obtained with M2 compared to M1 (smaller mSD
and higher mCC). The accuracies were also better with M2 (smaller
moffset).

4. Discussion
In this study, a new functional calibration procedure (M2) was
proposed to measure the 3D knee joint angle according to the ISB
recommendation. In contrary to previous calibration procedures in
literature, this method was suitable for a system only composed of
two IMUs. The performance of this new method was compared to a
simple method (M1) where visual alignment between IMUs and
segments axes was used.
The repeatability (w) of the functional calibration procedure
(M2) was comparable to that reported for calibration methods
relying on bony landmarks and using camera-based systems
(Della Croce et al., 1999, 2003). Unfortunately, none of the
studies proposing calibration procedures for MIMUs had reported
a comparable parameter (ODonovan et al., 2007; Picerno et al.,
2008).
The propagation of the calibration quaternions precision onto the
three knee joint angles was evaluated during gait using three

parameters (Table 1). Here, the comparison to literature is difcult

since the protocols and the parameters used to quantify the
repeatability are different among studies. Nevertheless, the functional
method (M2) proposed here seems to provide comparable results
than previous advanced calibration procedures based on standard
motion capture systems (Charlton et al., 2004; Hagemeister et al.,
2005; Donati et al., 2008). Regarding MIMUs, Picerno et al. (2008)
evaluated the repeatability of their anatomical-based calibration
method. However, they only propagated the calibration precisions to
the standing posture. In general, methods which use both anatomical
and functional calibration are reported to be more repeatable than
methods based only on anatomical calibration (Marin et al., 2003;
Schache et al., 2006). However, functional methods are sensitive to the
load applied to the joint (Marin et al., 2003; Mannel et al., 2004),
and to the execution of the calibration movements (Besier et al., 2003;
Marin et al., 2003). In order to minimize the inuence of the load
and to allow the methods to be used with more patients, the
calibration movements used in this study were passive. Passive FE
also minimizes the screw-home effect and limits the joint function
change respectively to the condition of the joint (Blankevoort et al.,
1988; Moglo and Shirazi-Adl, 2005), which should increase the
repeatability of the calibration procedure. Moreover, the passive FE
was carried out between 451 and 801 of exion and thus could be
assumed as pure knee joint FE (Blankevoort et al., 1990; Wilson et al.,
2000; Jan et al., 2002).
Finally, the three knee joint angles obtained through the proposed
inertial-based system and following both calibration methods
were compared to those of a reference system using three parameters
(Table 2). As expected, the errors obtained with M1 were higher than

ARTICLE IN PRESS
2334

J. Favre et al. / Journal of Biomechanics 42 (2009) 23302335

Ext > Flx,

60
40
20
0

Add > Abd,

20
0

20
0
20
40

Int > Ext,

20
10
0
10
20
Time, s
Fig. 4. 3D knee angles (FE, AA and IE) obtained during the walking task for a typical subject. The reference angles are in black and the angles obtained through the proposed
system in gray (dotted for M1 and solid for M2).

Table 2
Mean (SD) errors of the 3D knee joint angle.

FE
AA
IE

M1
M2
M1
M2
M1
M2

moffset

mSD

11.0
8.1
14.9
6.2
6.3
4.0

4.4
1.3
15.0
2.0
5.2
2.0

(5.1)
(5.4)
(3.5)
(5.1)
(4.3)
(4.7)

mCC
(1.4)
(0.5)
(2.6)
(0.6)
(1.7)
(0.9)

0.96
1.00
0.65
0.76
0.60
0.85

(0.02)
(0.00)
(0.16)
(0.18)
(0.26)
(0.11)

moffset and mSD are in degree.

those of M2 for every angles and parameters. Although both, M2 and

the reference, calibration methods dene the BAFs through several
functional and postural steps, they are based on different signals (i.e.,
angular velocity and acceleration vs. position and orientation).
Therefore, both BAFs calculations are not identical, and few
discrepancies between the angles were expected. Nevertheless, the
precisions (mSD) were very good in comparison to the repeatability of
previous methods using standard systems (Della Croce et al., 2005)
and better than the precisions reported by Picerno et al. (2008) which
used a wearable system. Moreover, the errors reported in Table 2
stand for the proposed and reference calibration methods and
therefore include the errors of both measurement systems (inertialand magnetic-based). These precisions, associated with the high
patterns similarity values (mCC), indicated a good agreement in the JCS
axes (e1e2e3) considered by both systems. The accuracies (moffset) were
much less satisfactory, which indicated differences in some BAF axes
that are not part of the JCS axes (Grood and Suntay, 1983; Besier et al.,
2003; Della Croce et al., 2003; Most et al., 2004; Favre et al., 2008).
This could be interpreted as a consequence of the lack of a denition
for the zero value of the 3D knee joint angle (Kadaba et al., 1990), and
the low accuracy (moffset) could largely be explained by the
discrepancies in the way to dene the zero angle values between
the proposed and reference systems.
As far as our knowledge is concerned, this was the rst time that
calibration procedures for IMUs were proposed and assessed to
describe knee joint function according to the ISB recommendation.

Visually aligning the IMUs with the segments (M1), as expected,

provided unsatisfactory results. In contrary, M2 showed a good
repeatability and agreement with the reference system. The accuracy
was limited and conrmed the caution required when comparing
absolute joint angles (Kadaba et al., 1990; Della Croce et al., 1999).
Passive movements were considered here because they are expected
to improve the repeatability. Nevertheless the procedure could be
adapted for active movements. Other studies including pathological
knee joints and follow-up over several months are in progress to
conrm the repeatability values and evaluate the relevance of this
system in clinical applications. Although this functional calibration
procedure was designed for IMUs, it can be used with MIMUs and
adapted to be used with standard motion capture systems. In
conclusion, the ambulatory system proposed in Favre et al. (2008),
which allows the measurement of the knee joint differential
orientation, combined with M2 provided a very easy-to-use tool to
record and calculate the 3D knee joint angle during short periods of
time. Through an exoskeleton harness was considered in this study,
the proposed combination can be used without harness if the soft
tissue artefacts are not critical (Leardini et al., 2005). Since this
measurement technique can be used outside a laboratory and in
unconstrained environments, it constitutes a very promising device
to monitor the knee joint angles during different kind of activities. A
future extension for the system could be the involvement of force
sensors and/or a musculoskeletal model in order to estimate joint
kinetics during daily activities. Finally, although this study focused
on the knee, the proposed technique could be used to monitor other
complex joints, such as ankle or hand, where adapted calibration
procedures would be necessary.

Conict of interest statement

The authors discloses any nancial or personal relationships
with other people or organisations that could inappropriately
inuence (bias) this work.

ARTICLE IN PRESS
J. Favre et al. / Journal of Biomechanics 42 (2009) 23302335

References
Andrews, J.G., 1984. On the Specication of joint congurations and motions.
Journal of Biomechanics 17, 155158.
Andriacchi, T.P., Alexander, E.J., 2000. Studies of human locomotion: past, present
and future. Journal of Biomechanics 33, 12171224.
Bachmann, E., 2000. Inertial and magnetic tracking of limb segment orientation for
inserting humans into synthetic environments. Ph.D. thesis, Naval Postgrduate
School, Monterey, CA, p. 176.
Baker, R., 2006. Gait analysis methods in rehabilitation. Journal of Neuroengineering and Rehabilitation 3, 4.
Besier, T.F., Sturnieks, D.L., Alderson, J.A., Lloyd, D.G., 2003. Repeatability of gait
data using a functional hip joint centre and a mean helical knee axis. Journal of
Biomechanics 36, 11591168.
Blankevoort, L., Huiskes, R., Delange, A., 1988. The envelope of passive knee-joint
motion. Journal of Biomechanics 21, 705709.
Blankevoort, L., Huiskes, R., Delange, A., 1990. Helical axes of passive knee-joint
motions. Journal of Biomechanics 23, 12191229.
Cappozzo, A., Catani, F., Croce, U.D., Leardini, A., 1995. Position and orientation inspace of bones during movement-anatomical frame denition and determination. Clinical Biomechanics 10, 171178.
Cappozzo, A., Della Croce, U., Leardini, A., Chiari, L., 2005. Human movement
analysis using stereophotogrammetryPart 1: theoretical background. Gait
and Posture 21, 186196.
Charlton, I.W., Tate, P., Smyth, P., Roren, L., 2004. Repeatability of an optimised lower
body model. Gait and Posture 20, 213221.
Della Croce, U., Cappozzo, A., Kerrigan, D.C., 1999. Pelvis and lower limb anatomical
landmark calibration precision and its propagation to bone geometry and joint
angles. Medical and Biological Engineering and Computing 37, 155161.
Della Croce, U., Camomilla, V., Leardini, A., Cappozzo, A., 2003. Femoral anatomical
frame: assessment of various denitions. Medical Engineering and Physics
25, 425431.
Della Croce, U., Leardini, A., Chiari, L., Cappozzo, A., 2005. Human movement
analysis using stereophotogrammetryPart 4: assessment of anatomical
landmark misplacement and its effects on joint kinematics. Gait and Posture
21, 226237.
Donati, M., Camomilla, V., Vannozzi, G., Cappozzo, A., 2008. Anatomical frame
identication and reconstruction for repeatable lower limb joint kinematics
estimates. Journal of Biomechanics 41, 22192226.
Favre, J., Jolles, B.M., Siegrist, O., Aminian, K., 2006. Quaternion-based fusion of
gyroscopes and accelerometers to improve 3D angle measurement. Electronics
Letters 42, 612614.
Favre, J., Jolles, B.M., Aissaoui, R., Aminian, K., 2008. Ambulatory measurement of
3D knee joint angle. Journal of Biomechanics 41, 10291035.
Foxlin, E., 1996. Inertial head-tracker sensor fusion by a complementary separatebias Kalman lter. In: Proceedings of the VRAIS, Santa Clara, CA,
pp. 185194.
Ganjikia, S., Duval, N., Yahia, L., de Guise, J., 2000. Three-dimensional knee analyzer
validation by simple uoroscopic study. Knee 7, 221231.
Grood, E.S., Suntay, W.J., 1983. A joint coordinate system for the clinical description
of 3-Dimensional motions-application to the knee. Journal of Biomechanical
EngineeringTransactions of the ASME 105, 136144.
Hagemeister, N., Parent, G., Van de Putte, M., St-Onge, N., Duval, N., de Guise, J.,
2005. A reproducible method for studying three-dimensional knee
kinematics. Journal of Biomechanics 38, 19261931.

2335

Jan, S.V., Salvia, P., Hilal, I., Sholukha, V., Rooze, M., Clapworthy, G., 2002.
Registration of 6-DOFs electrogoniometry and CT medical imaging for 3D joint
modeling.
Journal of Biomechanics 35, 14751484.
Kadaba, M.P., Ramakrishnan, H.K., Wootten, M.E., Gainey, J., Gorton, G., Cochran,
G.V.B., 1989. Repeatability of kinematic, kinetic, and electromyographic data in
normal adult gait. Journal of Orthopaedic Research 7, 849860.
Kadaba, M.P., Ramakrishnan, H.K., Wootten, M.E., 1990. Measurement of lowerextremity kinematics during level walking. Journal of Orthopaedic Research 8,
383392.
Kuipers, J.B., 2002. Quaternions and Rotation Sequences. Princeton University
Press, Priceton, NJ, p. 400.
Kutner, M.H., Nachtsheim, C.J., Neter, J., Li, W., 2005. Applied Linear Statistical
Models, fth ed McGraw-Hill, Irwin, USA, p. 1424.
Leardini, A., Chiari, L., Della Croce, U., Cappozzo, A., 2005. Human movement
analysis using stereophotogrammetryPart 3, soft tissue artifact
assessment and compensation. Gait and Posture 21, 212225.
Luinge, H.J., Veltink, P.H., 2005. Measuring orientation of human body segments
using miniature gyroscopes and accelerometers. Medical and Biological
Engineering and Computing 43, 273282.
Mannel, H., Marin, F., Claes, L., Durselen, L., 2004. Establishment of a knee-joint
coordinate system from helical axes analysisa kinematic approach without
anatomical referencing. IEEE Transactions on Biomedical Engineering
51, 13411347.
Marin, F., Mannel, H., Claes, L., Durselen, L., 2003. Correction of axis misalignment
in the analysis of knee rotations. Human Movement Science 22, 285296.
Mills, P.M., Morrison, S., Lloyd, D.G., Barrett, R.S., 2007. Repeatability of 3D gait
kinematics obtained from an electromagnetic tracking system during treadmill
locomotion. Journal of Biomechanics 40, 15041511.
Moglo, K.E., Shirazi-Adl, A., 2005. Cruciate coupling and screw-home mechanism in
passive knee joint during extension-exion. Journal of Biomechanics
38, 10751083.
Most, E., Axe, J., Rubash, H., Li, G., 2004. Sensitivity of the knee joint kinematics
calculation to selection of exion axes. Journal of Biomechanics 37, 17431748.
ODonovan, K.J., Kamnik, R., OKeeffe, D.T., Lyons, G.M., 2007. An inertial and
magnetic sensor based technique for joint angle measurement. Journal of
Biomechanics 40, 26042611.
Picerno, P., Cereatti, A., Cappozzo, A., 2008. Joint kinematics estimate using
wearable inertial and magnetic sensing modules. Gait and Posture
28, 508595.
Roetenberg, D., 2006. Inertial and Magnetic Sensing of Human Motion. Twente
University, Enschede, NL, p. 126 (Ph.D. thesis).
Schache, A.G., Baker, R., Lamoreux, L.W., 2006. Dening the knee joint exionextension axis for purposes of quantitative gait analysis: an evaluation of
methods. Gait and Posture 24, 100109.
Spoor, C.W., Veldpaus, F.E., 1980. Rigid body motion calculated from spatial
coordinates of markers. Journal of Biomechanics 13, 391393.
Sutherland, D.H., 2002. The evolution of clinical gait analysis: Part II Kinematics.
Gait and Posture 16, 159179.
Wilson, D.R., Feikes, J.D., Zavatsky, A.B., OConnor, J.J., 2000. The components of
passive knee movement are coupled to exion angle. Journal of Biomechanics
33, 465473.
Woltring, H.J., 1994. 3-D attitude representation of human jointsa standardization proposal. Journal of Biomechanics 27, 13991414.