Note: Descriptions are shown in the official language in which they were submitted.
CA 02595167 2007-07-30
METHOD FOR CALIBRATING SENSOR POSITIONS IN A HUMAN MOVEMENT
MEASUREMENT AND ANALYSIS SYSTEM
CROSS REFERENCE TO RELATED U.S PATENT APPLICATIONS
This patent application relates to U.S. provisional patent application Serial
No.
60/834,158 filed on July 31, 2006 entitled METHOD FOR CALIBRATING SENSOR
POSITIONS IN A HUMAN MOVEMENT MEASUREMENT AND ANALYSIS
SYSTEM, filed in English, which is incorporated herein in its entirety by
reference.
FIELD OF INVENTION
The present invention relates generally to a method for determining the six
fixed independent coordinates (3 displacements and 3 rotations) of body
surface
mounted motion sensors relative to the underlying skeletal frame, for the
purpose of
capturing six degreeof freedom (6-DOF) kinematic and kinetic information of
human
skeletal motion, and for analyzing the information in an anatomically and
physiologically meaningful way.
BACKGROUND OF THE INVENTION
Human movement analysis began formally at the end of the 19th century with
the advent of cinematography, and its application to capturing animal motion
by
pioneers such as Eadweard Muybridge (1830-1904) and Etienne-Jules Marey
(1830-1904). Unlike early motion capture systems, modern video and
optoelectric
human movement capture systems are accurate, reliable and fast, and have
applications spanning the clinical and biomedical sciences, sport sciences and
entertainment industries. While the goals of these different fields of
application may
vary considerably: eg. natural looking or contrived motion for entertainment
industry
versus accurate and objective motion data for clinical assessment, the
underlying
principles of motion tracking apply equally to all fields.
The preferred method of tracking any multi-segmented structure (such as a
human or animal) is to track the six degree of freedom (6-DOF) kinematics of
each
segment independently. There are numerous ways this can be accomplished, as
taught by those skilled in the art. A cluster comprising three or more light
reflective
markers or light emitting diodes can be placed on the skin of body segments,
or
1
CA 02595167 2007-07-30
placed on rigid plates which are then attached to body segments, and their 3D
positions tracked in space by video or optoelectric cameras. Another approach
is to
place magnetic field sensors on each body segment within an induced magnetic
field. Another approach again is to use microelectronic "MEMS" motion sensors,
such as accelerometers and gyroscopes, to estimate the 6-DOF kinematics of the
body segments (3-DOF plus a model). These technologies all fall into the
category
of surface mounted sensor systems ("sensor system") capable of measuring
directly
or indirectly 6-DOF kinematics.
However, to obtain both physiologically meaningful and clinically useful data
describing human movements, one needs to track the 6-DOF motion of the
underlying skeleton. Given that we are currently limited to surface mounted
technologies, we are forced to track the skeleton by inference, and as such,
we use
the surface mounted sensors to infer the underlying skeletal motions. As known
to
those skilled in the art, this is accomplished using a mathematical
"transformation"
that translates the 6-DOF sensor information into 6-DOF skeletal movements.
This
requirement is independent of the 6-DOF system selected for tracking skeletal
motion.
Review of Related Art
While different sensor systems have unique artifacts and sources of error,
virtually all body surface mounted technologies suffer from errors due to soft
tissue
movement. Various approaches have been taken to compensate for this naturally
occurring artifact, as disclosed in Lucchetti L, Cappozzo A, Cappello A, Della
Croce
U. Skin movement artefact assessment and compensation in the estimation of
knee-joint kinematics. J Biomech. 1998 Nov;31(11):977-84, and Cereatti A,
Della
Croce U, Cappozzo A. Reconstruction of skeletal movement using skin
markers: comparative assessment of bone pose estimators. J
Neuroengineering Rehabil. 2006 Mar 23;3:7. These teachings suggest that
optimal
mounting of sensors (or sensor arrays) is critical to minimize skin movement
artifacts. For the purpose of describing this invention it is assumed that
those skilled
in the art would employ such optimal mounting techniques to reduce skin
movement
artifact introduced into the surface mounted sensors' measurements. Thus we
2
CA 02595167 2007-07-30
continue our discussion focusing on the basic mathematical transformation
between
sensor and skeletal systems.
An approach to quantifying these transformations was disclosed in Riley PO,
Mann RW, Hodge WA. Modelling of the biomechanics of posture and balance. J
Biomech. 1990;23(5):503-6. for use with a camera system that tracks clusters
(or
arrays) of markers on rigid plates secured to the body segments. A method
employing a set of hand-held 6-DOF "pointer" arrays was used during a static
standing trial (subject stands perfectly still in a controlled posture) to
reference the
sensor system to the skeletal system for each body segment.
Others, such as Cappozzo A, Cappello A, Della Croce U, Pensalfini F.
Surface-marker cluster design criteria for 3-D bone movement reconstruction.
IEEE Trans Biomed Eng. 1997 Dec;44(12):1165-74, and Andriacchi TP, Alexander
EJ, Toney MK, Dyrby C, Sum J. A point cluster method for in vivo motion
analysis: applied to a study of knee kinematics. J Biomech Engng.
1998; 120:743-749 have taught how to acquire these transformations for
clusters of
skin mounted markers (placed upon various anatomical landmarks) as well. It is
worth noting that considerable effort has been taken to develop reliable
models of
skeletal motion when markers are placed directly on the skin as "deformable"
clusters. The deformability is assumed related to skin motion artifact and
thus can be
predicted and removed to improve joint center estimates, as taught by Lu T-W,
O'Connor JJ. Bone position estimation from skin marker co-ordinates using
global optimisation with joint constraints. J Biomech. 1999;32;129-134 and
validated by Roux E, Bouilland S, Godillon-Maquinghen A.-P, Bouttens D.
Evaluation of the global optimisation method within the upper limb kinematics
analysis. J Biomech. 2002;35:1279-1283; and further refined by Reinbolt JA,
Schutte JF, Fregly BJ, Koh BI, Haftka RT, George AD, Mitchell KH.
Determination
of patient-specific multi-joint kinematic models through two-level
optimization.
J Biomech. 2005;38:621-626.
But it is also worth noting that this form of "sensor artifact" is not due to
sensor slippage, since the skin mounted markers can only wobble, not slip,
relative
to one another. Wearable sensors for remote monitoring have the added
disadvantage of slippage in addition to skin related "wobble", and thus a
different
3
CA 02595167 2007-07-30
approach is needed to more generally tackle the problem. But independent of
the
approach taken the overall mathematical step is the same: to determine the
position
and orientation, or "pose", of the sensor coordinate system with respect to
the
skeletal coordinate system. As discussed above, the relative pose of one
system
with respect to another is generally expressed mathematically as a
transformation
matrix. Because this step is essentially a calibration step, we can also refer
to this
matrix as a calibration matrix.
Once a calibration matrix has been determined for each sensor and skeletal
segment, skeletal segments' poses in space are easily determined from the
sensors'
poses in space during arbitrary human postures and movements. Once the
skeletal
segments' poses are known in space, basic relative motion principles for rigid
bodies
known to those skilled in the art can be applied to compute the locations of
joint
centers between adjacent skeletal segments. For example, Riley et al (1990)
teach
how to use a chair rise trial (employing a linear range of motion of the knee
and hip)
to establish knee and hip joint centers of rotation in the sagittal plane. The
joint
centers describe the point or axis about which one skeletal segment rotates
with
respect to the other, and are of great interest in the field of movement
science for
both modeling and analysis of human movement, as well as for clinically
relevant
applications such as monitoring or diagnosing joint injuries or degenerative
joint
diseases.
The technique cited above for locating joint centers, as well as other
techniques published by those skilled in the art, may be applied, in most
circumstances, to any 6-DOF sensor system. The general limitation of the above
calibration approach, citing again Riley as an example, is that a separate set
of
instrumentation (eg. hand-held pointers or instrumented calibration frame) is
often
required to gather the data necessary to compute the sensor-to-segment
transformations. Requiring a separate set of calibration instruments may not
be
suitable for remote motion sensory systems applied in non-laboratory (real
world)
human body segment tracking applications.
Other calibration procedures, such as those that rely on precise positioning
of
skin markers on anatomical landmarks, may also not be suitable for extended
wear
motion tracking with MEMS type sensors. Without accurate instrumentation to
4
CA 02595167 2007-07-30
perform this calibration approach, and a general inability to control the
position of the
sensors on the body when applied in minimally or unsupervised environments,
estimation of the calibration matrices may be subject to considerable error.
These
errors result in non-physiologic skeletal motions, causing for example
adjacent
bones to unnaturally distract or impinge when they move relative to one
another.
In addition, should one of the body mounted sensors shift in its position
relative to the underlying skeleton ("slippage"), current teachings suggest
that the
calibration instruments must then be used again to re-establish the
calibration
matrix. This of course assumes that the sensor slippage is actually detected
during
data collection, which based on current teachings must be done visually.
US Patent 5,316,017 issued in 1994 discloses a glove having double-axis
sensors in the form of traducers to measure joint movements.
US Patent 5,533,531 issued in 1996 discloses a method for electronically
aligning a sensor having two nonparallel axes of measurement and being mounted
in a garment, so as to be positioned proximate to a joint's first and second
axes
respectively. The method involves a calibration step where one member on one
side
of the joint, the wrist for example, is held in a fixture while the other
member is
moved and initial calibration measurements are taken.
US Patent 5,791,351 issued in 1998 describes a motion measuring
apparatus using potentiometers connected together by mechanical linkages.
Rotary
potentiometers are attached to the joints of the wearers and calibration
consists of
physically aligning each sensor along the axis of rotation of the respective
joint
(column 5, lines 10-13).
US Patent 5,826,578 issued in 1998 is a parent application of US Patent
5,791,351 mentioned above and discloses basically the same information.
US Patent 6,050,962 issued in 2000 discloses a device for measuring the
joint angle of an articulated body. The sensors used are of the elongated
resisting
bend type, providing a voltage that is proportional to the alignment of their
ends. The
sensors are thin, flexible strips that include two variable-resistance
elements. The
sensors measure bone-to-bone angular orientation. Matrix manipulation and
iterative
approach are used to determine the position and orientation of one end of an
5
CA 02595167 2007-07-30
articulated mechanism assembly with respect to the other. (See column 9, lines
35-
42 and associated text).
US Patent 6,428,490 issued in 2002 is a continuation of the US Patent
6,050,962 mentioned above.
US Patent 6,127,672 issued in 2000 discloses a motion measuring device,
commonly referred to as a shape tape and in column 6, lines 53-56, "This shape
measuring tool may be coupled over all or part of its extent by constraining
means to
a portion of a body or object, the location, shape or orientation in space of
which is
to be measured." (column 6, lines 63-68) " It is sufficient for at least one
portion of
the sensor to be attached to a body for the location and orientation of that
portion of
the body to be determined with respect to a reference to a reference point
elsewhere
on the sensor." (column 7, lines 7-24) "Every sensor's location, and
orientation, can
be determined with respect to other sensors by inter-referencing the positions
of the
intervening sensors.
US Patent 6,692,447, issued in 2004 discloses a method to determine the
position of the knee joint and the hip joint on a person, using an optical
marker which
is affixed to the tibia of that person, and a camera. The leg with the marker
affixed to
it is moved in a pedaling motion and positions of the marker are recorded.
US Patent 6,997,882, issued in 2006 discloses a device and a method for
acquiring 6-DOF data regarding a person's movement, position, and orientation
in
three-dimensional space. The document describes a method to calibrate and re-
calibrate two accelerometer sensors relative to a reference Cartesian frame.
US Patent Publication Application No. 2005/0143676, published in 2005.
discloses a method for calibrating a device for studying knee kinematics. In
this
method, a first marker is mounted on the femoral portion of the leg and a
second
marker is mounted on the tibial portion of the leg. The leg is moved in a
kicking
motion and the position and orientation of the markers are digitized. A
position
calculator 42 is used to determine the axis of the knee. (Paragraph 0042).
CA Patent No. 1,208,747, issued in 1986 discloses a system for calibrating
the space coordinates of a robot gripper in six degrees of freedom. This
document
explains that errors in positioning the robot's gripper may occur due to drift
in some
of the six coordinate directions. Therefore compensation of the robot
coordinates at
6
CA 02595167 2007-07-30
suitable intervals is a requisite. The method disclosed includes moving a
gripped
object in a fixture having several sensors. The gripper is moved repeatedly
until an
error in the sensor readings is canceled.
CA Patent Application Serial No. 2,234,537, published in 1995 discloses a
range-of-motion-arm that has 6 degrees of freedom. The arm is capable of
measuring the movement of one body member relative to a fixed attachment on
another part of the body.
CA Patent Application Serial No. 2,246,290, published in 1997 discloses a
system to determine the location of a probe inside a patient's body. The
system uses
three field transducers. 'The relative position of the field transducers are
redetermined periodically and the position of the probe is redetermined
periodically
based on the redetermined relative position of the field transducers. This
system
permits the mounting of the field transducers of movable elements of the body,
as
for example, on the surface of the abdomen or thorax. Although this document
describes a step of calibration at interval to compensate for sensor movement,
it
does not suggest or describe the present method for calibrating sensor
positions.
CA Patent Application Serial No. 2,427,186, published in 2001 describes a
similar device as in the US Patent Publication 200510143676, mentioned above.
The document describes a harness for supporting three sensors on the femur of
a
person and one attachment bar for mounting another sensor to the tibia of the
person. The system provides position and orientation information of the femur
and
tibia in space, and the position and orientation of the sensors with respect
to one
another. The location of the sensors is detected at specific time intervals.
The present invention will address these limitations. We anticipate that the
teachings of this invention will be of interest to those developing and
utilizing
wearable sensor systems (and motion lab sensor systems as well) for human
motion
tracking.
SUMMARY OF THE INVENTION
The present invention addresses the above identified drawbacks by providing
a method that is capable of robust, accurate, and reliable location of
skeletal
coordinates systems with respect to the segment mounted sensors without the
need
7
CA 02595167 2007-07-30
for additional instrumentation, referred to herein as the "method". The method
can
be used in the laboratory setting in conjunction with the teachings of other
others in
the field, but more importantly will enable extended wear of wireless remote 6-
DOF
skeletal movement capture of a human or animal. Fields of application of this
technology may include, but are not limited to physical therapy rehabilitation
services; laboratory and field (remote) human movement science; high
performance
athletics; military training and simulation; animation library development;
advanced
gaming.
The method comprises a protocol for mounting of the sensor devices, a set of
initial calibration tests, and an analytical approach that utilizes these data
for arriving
at the sensor-to-segment transformations. The method acquires data from the
body
mounted motion sensor devices as taught by others in the field. Data from the
devices is then processed according to the present teachings to arrive at an
accurate representation of skeletal motion that can then be used for
sophisticated
biomechanical analyses and simulations, or used to build scaleable animation
libraries of complicated human movements, to list but a few examples.
An embodiment of the invention provides a method for locating a system of
motion sensors on a human or other animal body for capture of movement time
history, comprising the steps of:
determining mathematical transformation matrices between coordinate
systems of sensors strategically positioned on a person's body and skeletal
coordinate systems to produce a human body kinematic model;
refining the transformation matrices of the human body kinematic model to
minimize anatomical joint gap error; and
monitoring the anatomical joint gap to detect and correct for sensor slippage
during data collection.
A further understanding of the functional and advantageous aspects of the
invention can be realized by reference to the following detailed description
and
drawings.
8
CA 02595167 2007-07-30
BRIEF DESCRIPTION OF DRAWINGS
The aforementioned features and advantages, and other features and
aspects of the present invention, will be understood with reference to the
following
and accompanying drawings; wherein:
FIG. 1 illustrates a schematic block diagram of the overall method;
FIG. 2 illustrates a schematic block diagram of the method for mounting the
sensor system on the human body;
FIG. 3 illustrates a schematic block diagram of the sensor system calibration
stage;
FIG. 4 illustrates a schematic representation of the sensor-to-segment
coordinate systems used in the calibration stage;
FIG. 5 illustrates a schematic representation of the sensor-to-segment
transformations of the upper extremity;
FIG. 6 illustrates a schematic representation of the dynamic calibration stage
for optimizing the human body model;
FIG. 7 illustrates a schematic block diagram of the data capture and
processing stage; and
FIG. 8 illustrates a schematic block diagram of the sensor slippage monitoring
and correction stage.
DETAILED DESCRIPTION OF THE INVENTION
Generally speaking, the systems described herein are directed to methods for
calibrating sensor positions in a human or animal movement measurement and
analysis system. As required, embodiments of the present invention are
disclosed
herein. However, the disclosed embodiments are merely exemplary, and it should
be
understood that the invention may be embodied in many various and alternative
forms. The Figures are not to scale and some features may be exaggerated or
minimized to show details of particular elements while related elements may
have
been eliminated to prevent obscuring novel aspects. Therefore, specific
structural
and functional details disclosed herein are not to be interpreted as limiting
but
merely as a basis for the claims and as a representative basis for teaching
one
skilled in the art to variously employ the present invention. For purposes of
teaching
9
CA 02595167 2007-07-30
and not limitation, the illustrated embodiments are directed to methods for
calibrating
sensor positions in a human movement measurement and analysis system.
As used herein, the term "about", when used in conjunction with ranges of
dimensions, angles or other physical properties or characteristics, is meant
to cover
slight variations that may exist in the upper and lower limits of the ranges
as to not
exclude embodiments with concentrations slightly above or below those recited
herein. It is not the intention to exclude embodiments such as these from the
present
invention.
The illustrative embodiment of the present invention provides method for the
calibration of body mounted sensors' positions and orientations relative to
the
skeletal frame, for analysis of kinematics and kinetics of human movement. The
method is independent of the means by which 6-DOF kinematic measurements of
body mounted sensors are acquired. Data captured by the system being employed
are used as inputs to the method of the present teachings. The method of the
present teachings utilizes a protocol for positioning the person for static
calibration of
the sensor-to-segment transformations, following by a protocol involving
dynamic
limb movements for fine-tuning the skeletal model. Once the sensor-to-segment
transformation are established, sensor data are collected and processed, as
taught
by others, to arrive at skeletal motions. The method presented herein also
teaches
how to monitor the sensor data to detect and compensate (correct) for sensor
slippage. The method may be used for both humans and animals but in the
following
description the method is exemplified with reference to the human body.
FIG. 1 is a schematic block diagram of the method according to the teachings
of the present invention. The present invention relies on the mounting of the
sensor
devices ("system") in step 2 upon a person in a precise and specific manner
consistent with the operation of the system's devices. This is followed by the
calibration step 4, where the kinematic model of the human body for the person
is
created. Once satisfactory model error tolerances have been reached, the
system is
used in live capture mode and skeletal kinematic data are computed and stored
in
step 6. Finally, the data acquired during the session are used to monitor and
correct
for sensor slippage in step 8.
CA 02595167 2007-07-30
FIG. 2 is a schematic block diagram of step 2, for mounting the devices of the
system on the person. First, a combination of sensors are selected that meets
the
needs of the task being monitored in step 16 of FIG. 1. The next step 12 of
FIG. 2 is
to acquire a set of anatomical measurements that will be used to develop the
skeletal model. These may consist of various anatomical measurements, selected
according to the teachings of others, such as Riley et al. (1990). The
garment(s)
and/or cuff(s) with the motion sensors are donned by the user and the system
is
powered up in step 14 of FIG. 2, as taught by others in the field.
Recall that step 4 of FIG. 1 showed the step consisting of the calibration
methods. This step is shown in more detail in the schematic block diagram of
FIG. 3.
Calibration commences with positioning the body in a known and controlled
posture
in step 16 of FIG. 3. This step requires the person wearing the sensors to
either
stand erect with feet spaced at a specific distance apart and arms hanging
vertically
or (for those with disabilities) to sit briefly in a special straight backed,
level seated
chair. In static calibration mode (standing or sitting), data are acquired
from the
sensors for a specified time in step 18 of FIG. 3, and used with the
anatomical data
to compute and estimate the positions of the skeleton relative to the
segmented
mounted sensors in step 20 of FIG. 3, see Data Processing step shown in Figure
7
for more details.
Once static calibration is complete, the person then proceeds (depending on
which body segments have sensors mounted on them) to a brief dynamic
calibration
session. In this step, the user first initializes the dynamic calibration
protocol in step
22 of FIG. 3, and then performs a series of range-of-motion trials to acquire
information about the relative movements of the body segments in step 24 of
FIG. 3.
These could consist of any or all of the following: for the arm: shoulder
abduction/adduction and flexion/extension, elbow flexion/extension, arm
pronation/supination (rotation of the forearm), and wrist flexion/extension;
for the leg:
hip abduction/adduction and flexion/extension, knee flexion/extension, and
ankle
dorsiflexion/plantarflexion. For the whole-body: in addition to above, trunk
flexion/extension and neck flexion/extension, etc.
The method taught with the present invention shows how these data are used
to fine-tune the skeletal model by computing and minimizing the gap at
skeletal
11
CA 02595167 2007-07-30
joints in step 26 of FIG. 3. This is accomplished by re-computing the sensor-
to-
segment transformations iteratively until the joint gaps are below a specified
threshold and measured segments lengths are maintained within a specific
threshold. This is based on the fact that computed joint centers should be
such that
skeletal segments do not distract or impinge beyond known physiologic limits.
FIGS.
4 to 6 show specific analytical steps used using the elbow as an example (it
will be
appreciated that the method disclosed herein may be used with any body
segments
connected by an anatomical joint).
As shown in FIG. 4, the long axis of the upper arm passes through center of
circles J, and J2 (biceps cross-section) ul at distance r, from the sensor S,
along
the -zSl axis.
u, =s1+[0,0,-r,][41]
The elbow J2 is located a distance Dy below ul along -yB1 axis
J2 =u1 +[0,-DY,O]
The shoulder J, is located a distance L, above J2 along the +yB1 axis
Jl =J2+[O,L,,0]
The orientation of the upper arm coordinate frame is assumed [0,0,0] degrees,
thus
its rotation matrix is an identity matrix.
1 0 0
~Bl = 0 1 0
0 0 1
Now the sensor-to-segment transformations for the upper arm can be computed.
The relative rotation matrix: esi =[OsI ~ [OB1 ~
Shoulder position in sensor coordinates: Psi =(JI -SI )~STI
Elbow position in sensor coordinates: Ps =(JZ -S1 )Oi
FIG. 5 (top) shows how shoulder and elbow are located relative to the upper
arm sensor Sl. Next the wrist center is located from the forearm sensor
located on
the dorsal wrist surface, as shown in FIG. 5(bottom).
J3 =sZ+[0,0,-rz]A2]
12
CA 02595167 2007-07-30
C,x CXY Cx~
Let 0B2 = CYX Cri CYZ
C~ CZY C~
The x-axis of the forearm system B2 is assumed to be co-linear with the x-axis
of the
sensor system S2. Therefore C. ={C., CXY, C,,,} is taken from the sensor
rotation
matrix OS2. The y-axis direction vector is found from
C = X Jz - X J3 yJz yJ3 ZJz - ZJs
Y
J2 - J3 ' J2 - J3 JZ - J3
And the z-axis direction vector can be located by the cross-product CZ = CX x
Cy
Now the sensor-to-segment transformations for the forearm can be
computed.
The relative rotation matrix: 9sZ =[OSZ ]T [OBZ ]
Elbow position in sensor coordinates: PS2 =(J2 - S2 )[OS2 ]T
Wrist position in sensor coordinates: PSZ =(J3 -SZ)[OSZ]T
Once the above sensor-to-segment transformations have been stored (in a
separate file or database, and/or written to a header of a data file), for any
arbitrary
trial (arm activity) the 6-DOF position and orientation of the skeletal body
segment
can be found, by inverse transformation using the sensor-to-segment
transformation
matrices.
For upper arm segment:
Shoulder position in global coordinates: J1 = S, +Ps 0sI
Elbow position in global coordinates: JZ'~ = S1 +Ps Os~
Rotation matrix of upper arm: OBl =OsIesi
For forearm segment:
Elbow position in global coordinates: Jz2' = Sz +Psz~sz
13
CA 02595167 2007-07-30
Wrist position in global coordinates: J3 =S2 +Ps Osz
Rotation matrix of forearm: OB2 = Oszesz
Once the orientation of the upper arm and forearm is found in 3D space, we
can then compute elbow angular displacements using the relative rotation
matrix
e,2 =[0B,][0BZ]T , and which is easily solved for flexion/extension a,
abduction/adduction R, and internal/external rotation y, angles. This can be
done
using a Cardan 3-1-2 decomposition of the rotation matrix, as embodied herein,
or
other matrix decomposition method known to those skilled in the art.
Joint Center Determination
As an illustrative example of one possible embodiment, consider the human
elbow joint. The elbow has essentially two rotational degrees of freedom:
flexion-
extension and internal-external rotation. Unfortunately (for the modeler),
this motion
is facilitated by the two forearm bones, which can move relative to one
another. The
elbow joint model is considerably simplified if we assume it behaves as a 2-
DOF
joint with axes of rotation passing through a fixed position on both segments.
This
requires that both fixed points on each segment are always coincident (the
center of
rotation) in space.
The 6-DOF tracking of upper arm and forearm gives us a convenient
opportunity to fine tune the anatomical model of the elbow. Our initial
measurements
for locating the elbow was only to put the model through its first iteration.
Surely
when we perform a range of motion task, the elbow on the upper arm JZ0) will
not
perfectly coincide with the elbow on the forearm J2(2) . The degree of
mismatch tells
us the degree we erred in finding the joint center of rotation from our simple
anatomical model.
We can improve this model, however, if we simply apply an iteration approach
to finding the anatomical model which closes the apparent "joint gap". The
procedure
is as follows:
Model Refinement
After calibration, the segment coordinate axes B and joint centers (J; and
J;+,)
can be expressed in Sensor S coordinates. For multiple connected segments,
each
14
CA 02595167 2007-07-30
with a calibrated sensor, the endpoints of each segment can be found in global
space. The time history of movement of the segments in global space is
illustrated in
FIG 6. It is the joint gap that must be minimized.
One way this might be done is to adjust the segment coordinate system B
relative to S to minimize the joint gap. Since the proximal and distal joint
centers
define the B long (y) axis, we are essentially just manipulating the segment
long
axes (we can redo the cross-products to get the modified x and z axes later)
to find
the best joint center location.
Assume we look at the relative movement of segment B2 with respect to B1.
The path of J2 on segment B2 (J2(2)) would trace a path relative to B1. It
would
coincide with the fixed center J2 on BI (J2(1)) only once - at neutral
position.
Now we simply average the path of J2(1) = J2(')' and this becomes the new
origin of B1' (and passing through J1). We then re-compute the segment B1
axes,
and compute their position and orientation relative to S1. Now we do the same
exercise, except for the distal segment's J2 location. Get average J2(2) =
J2(2)' to
define new B2 relative to S2. If we now re-run the analysis with the modified
calibration, our joint gap should be smaller.
The above can be run multiple times until the gap is minimized (for example,
when changes less than a specified threshold (eg. 1 mm) occur with each
additional
iteration).
Analytical Procedure
1. Collect range of motion data (after static trial is done and applied)
2. Using sensor-to-segment transformations:
a. Compute trajectory of elbow center on upper arm J2(1) in global coordinates
b. Compute trajectory of elbow center on forearm J2(2) in global coordinates
3. Compute the RMS distance e; between J2(') and J2(2)
4. Transform J2(2) into B, coordinates = J2(Z)B'
5. Find the mean of the excursion of J2(2)B': this becomes the new location of
J2('), on
Bi.
6. Transform J2(') into B2 coordinates = JZ(' )B2
CA 02595167 2007-07-30
7. Find the mean of the excursion of J2 (1)B2: this becomes the new location
of J2(2)' on
B2.
8. From J1 - J2(1), and J2(2)'- J3 re-compute the sensor-to-segment
transformations.
9. Repeat steps 2 and 3.
10. Compare ei to previous e;_1. If less than a set threshold (eg. 1 mm), no
further
improvement expected. If greater than a set threshold, steps 4-10 are
repeated.
11. Store the final sensor-to-segment transformations.
Now recall that step 6 of FIG. I shows the step consisting of active data
collection mode. This step 6 is shown in more detail in the schematic block
diagram
of FIG. 7. Once the device(s) are calibrated, the system is switched to active
mode
in step 28. Sampling rates and data storage protocol are determined by the
data
collection system being used, as taught by those in the field. During data
collection
mode the sensors capture data during tasks or activities of daily living as
desired by
the user in step 30, and the sensor data transformed into skeletal movement
kinematics using the stored sensor-to-segment calibration matrices in step 32.
Finally, the skeletal kinematics are stored by the system in step 34.
Finally recall that step 8 in FIG. 1 showed the step consisting of monitoring
and correcting for sensor slippage. This step is shown in more detail in the
schematic block diagram of FIG. 8. The first step 36 is to monitor the joint
gap
magnitudes of all joints being measured during the data collection trial. If
at any point
in time the joint gap magnitude exceeds a given tolerance for a given period
of time,
the method we taught above describing the dynamic calibration in step 24 of
FIG. 3
and FIGS. 4 to 6 is automatically initiated and corrections made to the
segment-to-
sensor transformations "on the fly" in step 38 of FIG. 8. If the joint gap
magnitude
does not improve to tolerances with the above step, the data collection trial
is halted
in step 40 of FIG. 8. At this point the calibration step 4 of FIG. 1 comprises
of steps
16 through 26 in FIG. 3 is re-initiated as required.
It is believed that this document complies with the requirements of 35 U.S.C.
112, as it provides sufficient information to enable those skilled in the art
to build and
use this invention.
16
CA 02595167 2007-07-30
Numerous modifications and alternative embodiments of the invention will be
apparent to those skilled in the art in view of the forgoing description.
Accordingly,
this description is illustrative only and is for the purpose of teaching those
skilled in
the art the best mode for carrying out the invention. Details of the structure
may vary
substantially without departing from the spirit of the invention, and
exclusive use of
all modifications that come within the scope of the appended claims is
reserved. It is
intended that the invention be limited only to the extent required by the
appended
claims and the applicable rules of law.
As used herein, the terms "comprises", "comprising", "including" and
"includes" are to be construed as being inclusive and open ended, and not
exclusive.
Specifically, when used in this specification including claims, the terms
"comprises",
"comprising", "including" and "includes" and variations thereof mean the
specified
features, steps or components are included. These terms are not to be
interpreted to
exclude the presence of other features, steps or components.
Bibliography
Lucchetti L, Cappozzo A, Cappello A, Della Croce U. Skin movement artefact
assessment and compensation in the estimation of knee-joint kinematics. J
Biomech. 1998 Nov;31(11):977-84.
Cereatti A, Della Croce U, Cappozzo A. Reconstruction of skeletal movement
using
skin markers: comparative assessment of bone pose estimators. J
Neuroengineering Rehabil. 2006 Mar 23;3:7.
Riley PO, Mann RW, Hodge WA. Modelling of the biomechanics of posture and
balance. J Biomech. 1990;23(5):503-6.
Cappozzo A, Cappello A, Della Croce U, Pensalfini F. Surface-marker cluster
design
criteria for 3-D bone movement reconstruction. IEEE Trans Biomed Eng. 1997
Dec;44(12):1165-74.
17
CA 02595167 2007-07-30
Andriacchi TP, Alexander EJ, Toney MK, Dyrby C, Sum J. A point cluster method
for
in vivo motion analysis: applied to a study of knee kinematics. J Biomech
Engng.
1998;120:743-749.
Lu T-W, O'Connor JJ. Bone position estimation from skin marker co-ordinates
using
global optimisation with joint constraints. J Biomech. 1999;32;129-134.
Roux E, Bouilland S, Godillon-Maquinghen A.-P, Bouttens D. Evaluation of the
global optimisation method within the upper limb kinematics analysis. J
Biomech.
2002;35:1279-1283.
Reinbolt JA, Schutte JF, Fregly BJ, Koh BI, Haftka RT, George AD, Mitchell KH.
Determination of patient-specific multi-joint kinematic models through two-
level
optimization. J Biomech. 2005;38:621-626.
Reinbolt JA, Haftka RT, Chmielewski TL, Fregly BJ. A computational framework
to
predict post-treatment outcome for gait-related disorders, Med Eng Phys. 2007,
In-
press: doi:10.1016/j.medengphy.2007.05.005
18
