You are on page 1of 13

COMPUTER GENERATION OF HUMAN GAIT

KINEMATICS*?

M. Y. ZARRUGH~ and C. W. RAIXLIFFE$


Department of Mechanical Engineering, University of Michigan.
Ann Arbor, MI 48109, U.S.A.
and
Biomechanics Laboratory, Department of Mechanical Engineering,
University of California. Berkeley, CA 94720, U.S.A.

Abstract -The paper describes a computer program that generates absolute motion variables of human gait
from predetermined relative motions. Relative displacements are measured over a range of step rates during
both free (self-determined step rate at different speeds) and forced (forced step rate at a constant speed)
walking, converted into harmonic coefficients and stored in an array as a function of step rate. Only six
variable identifiers need to be specified to compute any absolute variable or its derivatives at any desirabie
step rate. The paper displays some examples of measured relative motions and reconstituted absolute
variables.

lNTRODUCT’lON Measurements of absolute gait variables are best


accomplished by photography in its various forms.
Kinematic variables of gait are valuable to both Photographic techniques of motion measurements
dynamic synthesis and analysis in the studies ofhuman include interrupted light method, cinematography and
walking. In dynamic analysis, the measured kinematic optoelectronic recording.
data of body segments are used to predict forces and Marey (1895) invented the interrupted light mathod
moments applied to each isolated segment. Kinematic which, in its ever-improving form, has been used by
measurements are needed in synthesis problems for
authors in all of the classic studies on human gait
comparison to motions predicted by the synthesis (Carlet et al., University of California), as well as in
from a known or assumed set of moments and forces, more recent studies (Miiller and Hettinger, 1952b;
thus providing a means of evaluating the accuracy of Drillis, 1958 ; Murray et al., 1964; Milner et al., 1973).
the assumed model. The reader is referred to Glanville and Krewer’s (1937)
This paper describes a computer program which
comprehensive review of the literature on gait
generates absolute kinematic data from predetermined measurements prior to 1937.
relative motions. The relative displacements are mea-
Motion pictures, when Gk& &a moving body and
sured over a range of walking speeds, converted into
of a timer simultaneously, serve as a satisfactory record
harmonic coefficients and stored in an array as a
of absolute motions. This method is widely used
function of step rate. Only six variable identifiers need
because of the simplicity of the data gathering equip-
to be specified to compute any desired variable or its ment, ease of data collection and lack of encumbrance
derivatives at any desired step rate. The paper also to the test subject (University of California, 1947;
displays examples of the measured relative motions Levcns et al., 1948; Liberson, 1965; Sutherland and
an&reconstituted absolute variables. Hagy, 1972; Cappoao et al., 1975). During data
The number and variety of kinematic measurements
gathering mirrors are sometimes placed above the
of walking are endless, limited only by the facility and subject (Murray et al., 1967a) or below a glass walkway
ingenuity of the researcher. The choice of a particular
(Ryker, 1952) at a 45’ angle to obtain multiple views on
method of measurement, however, is dictated by the ‘the same photographs. Three dimensional displace-
purpose of the investigation which is reflected in such ments can also be recorded with stereophotography
considerations as the motions to be measured, a (Ayoub et ol., 1970). In a medical setting, motion
sampling rate sufficient to reproduce the desired picture photography with a stationary or a moving
degree of detail, accuracy of the measurements and camera is advantageous over other methods especially
ease of instrumentation and data reduction. when the data reduction is mechanized through the use
of a motion analyzer (Sutherland and Hagy, 1972).
* Recehed for publication 22 August 1978. Optoelectronic recording of motion, a relative,new-
t Research support through contract VAV 101 (134) with
the Veterans Administration. comer to gait studies, includes television recording and
$ Assistant Professor in Mechanical Engineering. photodetector plate methods. The first method is
9 Professor of Mechanical Engineering. closely related to cinematography, except that the
RM 12 ?--A
99
100 M. Y. ZARRUGH and C. W. RADCLIFFE

motions are recorded on Videotape instead of film and horizontal components (Liberson et al., 1962;
(Winter et al., 1947a and 1974b; Cheng, 1974). The Gage, 1%7), horizontal component (Waters et al.,
method becomes effective when sophisticated translat- 1973) and all three components including the lateral
ing equipment are used to reduce the data directly into (Cavagna et al., 1%3 ; Gersten et al., 1969) component
coordinates (Winter et al., 1972). The photodetector of the acceleration of a point on the trunk approximat-
plate method (Selspot), still in the test stage, converts ing the center of mass of the body to compute an
the images of points of interest directly into electrical estimate of the total force between the feet and the
signals indicating planar position (Lindholm, 1974; ground. Liberson (1965) measured angular accele
&erg, 1974). ration of the shank for correlation with phasic acti-
String transducers are a simple and direct means of vities of major hip joint muscles. Accelerometry studies
measuring absolute displacements on the treadmill. emphasize the frequency spectra of acceleration to
The data are obtained in finished form requiring a distinguish between normal and pathological gait
minimum of further reduction. The main disadvantage (Gage, 1967; Smidt et al., 1977). Morris (1973) pre-
is that as many strings as displacements must be sented a technique using a specially built strain gauge
attached to the subject. The type of motion measured accelerometer to describe the total motion of the
depends on what instrument is linked to the other end shank.
of the string. Linear displacements (Ralston and
Lukin, 1%9; Lamoreux, 1971; Waters et al., 1973), MODElJNC OF THE WAIJUNG MECHANISM

velocity (Drillis, 1958) and acceleration can be


measured. Structure
The body accomplishes its essentially forward dis- Analysis in this paper is based on a model of the
placement by means of a highly coordinated series of human skeleton consisting of seven articulated seg-
rotations of lower limb segments. There are two ments as suggested by Elftman (1967). It includes three
common methods of measuring body rotations in fundamental divisions: right and left lower limb linked
walking, goniometry and pin studies. A goniometer is together by a composite segment, the HAT
a device to measure relative angular rotation between (head-arms-trunk). Each lower limb is divided into a
two body segments. To obtain accurate measurement thigh, shank and foot-shoe combination. In all, the
of rotation about a joint, it is necessary to align model consists of 12 links, six of which comprise the
precisely the measurement axis with the anatomical HAT: pelvic girdle, thorax and head, right arm, right
axis and to insure that the goniometer does not restrict forearm, left arm and left forearm. Figure 1 shows a
the range of joint rotation. Lamoreux (1974) de- schematic representation of this model in which each
veloped the self-aligning goniometer in which a paral- segment is identified with a number starting at the
lelogram linkage spans a joint allowing rotations to be pelvis and proceeding distally on the right side and
measured without interference with the joint motion then the left. The assumption of symmetry between
and without the need for axis alignment. He used the right and left sides permits the determination of gait
method extensively in his complete and systematic variables of the left side from those of the right side by
kinematic study of gait (1970, 1971). Other studies appropriate shift along the time axis.
presented segmented goniometric measurements of a
Mass and Geometric Parameters
few joints with little attention to the effect of speed or
axis alignment (Karpovitch et ol., 1959 ; Gollnick et al., Despite voluminous published anthropometric
1964; Wright et al., 1964; Finley et al., 1969; Johnston data, those of Braune and Fischer (1889) remain the
and Smidt, 1969; Chao et al., 1970; Kettelkamp et ol., most useful and complete. They expressed anthropo-
1970; Smidt 1971; James and &erg, 1973 ; Milner and metric parameters as ratios or coefficients, so that
Dali, 1973). Townsend et a/. (1977) described a six-
degree of freedom goniometer to measure the total
motion of the knee. n
When it is very difficult to attach a measuring
instrument onto a segment (e.g. the vertebral column),
or more accuracy is needed, pins are inserted directly
into the bone, mechanically linking it to a measuring
device (Close and Inman, j953 ; Kinzel et al., 1972b) or
to prominent targets for photography (Levens et al.,
1948).
Measurements of velocity and acceleration are
useful for comparison to derivatives computed from
measured displacements. Numerical differentiation,
without smoothing or filtering, preferentially mag-
nifies high frequency noise, but careful filtering (Winter
et al., 1974a ; Pezzack et al., 1977) overcomes the
problem. Many investigators measured the vertical Fig. 1. Schematic representation of body structure.
Computer generation of human gait kinematics 101

Table 1. Approximate coordinates of the centers of mass and joints in the reference
position

Coordinates relative to height

Segment Mass Horizontal Vertical Lateral


Segment No. ratio x ! .’

HAT* 0 0.680 0.0 0.644 0.0


Pelvis 1 0.278 0.0 0.560 0.0
Thigh 2 0.100 - 0.0024 0.420 0.047
Shank 3 0.045 -0.0016 0.178 0.043
Foot 4 0.015 0.0100 0.018 0.05 1
Hip - - 0.0 0.527 0.049
Knee - - 0.0 0.300 0.045
Ankle - - -0.1070 0.042 0.041

l Includes head, arms and the trunk.

total body mass without shoes. The personal data of


the subject of this study are tabulated in Table 2.

RELATIVE DISPLACEMENT MEASUREMENTS

Instrumentation and data gathering


The system of measurements (Fig. 3) as designed by
Lamoreux (1970), is used to locate the pelvic girdle in
space via its six degrees of freedom and then to
measure the rotations of each lower limb segment
relative to the one proximal to it, starting with
measurements of thigh rotations relative to the pelvic
girdle. Locating the pelvis in space is accomplished by
three pairs of parallel strings connecting the pelvic
girdle to spring-loaded instrumented arms (Fig. 2)
above, behind and to the right of the subject. For the
sake of simplicity, only the rear arm, measuring the
Fig. 2. Definition of the coordinate system. forward displacement of the pelvis and its rotation
about a vertical axis, is shown in Fig. 2. Joint rotations
are measured by potentiometers (Cummings and
Lamoreux, 1972) attached to segments by means of
differences between individuals were taken into ac-
count. Their data on center of mass and approximate
joint locations are used in this study. As tabulated in
Table 1, these three dimensional coordinates (see Fig. 2
for a definition of the coordinate system) are expressed
as ratios of height without shoes. Only parameters of
the right limb are shown, because left to right sym
metry is assumed. Parameters expressing the distri-
bution of mass in the body are obtained from Braune
an.d Fischer’s data or computed from geometrical
approximations of segments as truncated ellipsoidal
cones, using Dempster’s data (1955) for establishing
the major and minor axes of the conical cross-sections.
Table 1 lists the ratio of segment masses relative to

Table 2. Subject’s personal data

Age, yr 54
Body mass, barefoot, no instruments, kg 70.1
Body mass with shoes and instruments, kg 72.2
Shoes mass, kg 1.5
Height above floor without shoes, cm 176.8
Shoe heel height. cm 2.5
Shoe length, cm 30.0
Fig. 3. System of motion measuremem.
102 M. Y. ZARRUGHand C. W. RADCLIFFE

light aluminum frames and spanning the joints with RESULTS


self-aligning linkages (Lamoreux, 1974). In addition to
the six displacements of the pelvis, hip, knee and ankle Step length and step rate
flexion angles are measured. Hip adduction is in- This basic gait pattern has been shown (Zarrugh et
terpolated from Lamoreux’s (1970) data to match the al., 1974) to .be highly related to metabolic energy
current set of measurements. Data during sixteen full expenditure. Its influence on the kinematic aspects of
walking cycles measured from right heel contact to the gait are discussed elsewhere by Lamoreux and Todd
next right heel contact, are sampled at a rate of 200 (1974). They showed that kinematic measurements on
samples per set and stored in digital form. The heel the same subject are highly repeatable if they were
contact marks are generated by microswitches at- made under the same conditions of step length and
tached to the heels of both right and left shoes. The heel frequency. However, varying the speed, step rate (n) or
events data as well as the treadmill displacement in step length (s) had a profound effect on the measured
millimeters are also recorded. motions. Therefore, it is necessary to accompany any
Kinematic data during both free (self-selected step measured kinematic data by the associated data on
rate) and forced (forced step rate) walking are col- average walking velocity, step rate and step length.
lected. The free pattern data are gathered over a range The s-n pattern data for our subject are displayed in
of treadmill speeds corresponding to a step range of Fig. 4 in which the solid line depicts the constant s to n
8%130steps/min at 5 steps/min intervals. In the forced ratio fit and the interrupted line displays the constant
pattern experiment, the step rate was forced on the slope relationship (s = c, + cln). A comparison be-
subject via an electronic metronome within a step rate tween these two forms in terms of the sum of squared
range of 90-140 steps/min at a constant speed of 90 errors shows a slightly smaller sum (38.5cm’) or
m/mitt. To overcome subject dependent random vari- slightly better fit, for the constant s-n ratio form than
ations, all gathered data are averaged over sixteen the constant slope relationship which has an error sum
cycles after reviewing for consistent features. Data of 50.3 cm2.
processing and storage are facilitated by compressing
the data into harmonic coefficients form. A fast Free pattern kinematic data
Fourier transform program converted all measured Each of the measured kinematic displacements is
variables into harmonic coefficients. Seven to twelve presented as a family of curves, one for each tested step
harmonics are used to reconstruct all gait variables, rate (or speed), as a function of time in percent of cycle
except ankle flexion which required at least 20 har- to facilitate visualization ofspeed effect (Fig. Sa-i). The
monics for proper reconstitution. corresponding harmonic coefficients are available

STEP FREQUENCY, STLPS/MIN

n,STEP FREPUENCY, STEPS/SEC

Fig. 4. Step length, step rate and velocity measures.


PELVIS POSITION PELVIS ROTATIONS FLEXION ANGLES

* 1

I . , , . I * , J . , f
50 1 T-- 100 1
50
Time. “/- Of cycle Tirne59/. of cycle Time. % of cycle

Fig. 5. Measured displacementsin free walking.


PELVIS POSITION PELVIS ROTATION’ FLEXION ANGLES
I

I * I I I I I , I I I , ,
0 50 100
Time. % of cycle Time, cycle Time, ‘/a of cycle

Fig. 6. Measured displacements in forced walking.


Computer generation of human gait kinematics 105

from the authors. The hip adduction angle data were Forced pattern data
interpolated from Lamoreux (1970).
One of the remarkable features displayed in this In order to investigate the conditions under which
data is the abrupt change in the flexion patterns of the the body minimizes total work done during loco-
hip, knee and ankle joints occurring at about 105 motion, it is necessary to force combinations of step
steps/min. Below this transitional step rate, in the length and step rate which are not adopted normally in
range of 85-100 steps/min, the flexion patterns for gait. This is accomplished by keeping the speed
each joint are almost identical showing almost no constant at 90 m/min, while imposing a set of varying
effect due to increased step rate. Within the same step step rates between 90 and 140 steps/min. Figures 6a-c
rate range, the knee remains almost fully extended display the measured kinematic variabbs for this
(Fig. 5h) during most of stance phase ((r-40”/,). Maxi- pattern of walking. The harmonic coefficients for the
mum ankle extension which normally occurs shortly same data are available from the authors.
(less than 1% cycle) before toe-off is observed in Fig. 5i Despite constant speed, an observable effect is
to take place essentially at the same time (66% cycle). shown of the changing step length on each measured
Increased step rates above the transitional one result in variable. This dramatically illustrates how the basic
increased knee flexion during stance phase (Fig. 5h), gait parameters, speed, step length and step rate,
earlier and earlier toe-ofTpoints (Fig. 5i) and increased influence the kinematic variables of walking. Lower
maximum hip flexion (Fig. Sg). limb flexion angles (Figs. 6a-i) are not greatly affected

USER SWPLIED IDERTIFIERS

l SEGLIENT (ISEG) @WERATE RELATIVE MOTION


.VARIABLE (IVAR)
DATA AS k FWCTIIIN OF

aMuLTIPLE SEGmlT (WLTI) TIME IN PERCENT OF CYCLE


. STEP RATE (CADW) 7
1

CURRENT SEGMENT = 1
STDRLDHaRIow1c
CDEFFICIEWS~
I

PERFDRII VELOCITY ANALYSIS ON CUR-

PERFDRM ACCELERATION ANALYSIS ON


CuRRfNT SEGNENT

Fig. 7. Simplified flow chart of the method of solution for absolute motions from measured displacements.
106 M. Y. ZARRUGH
and C. W. RADCLIFFE

by change in step length. In contrast, the step length measured successively at the current orientation of the
has the largest influence on pelvic displacements, instrument axes (Fig. 12). The final configuration of
especially on the three linear displacements (Figs. any line on the link is determined in terms of a rotation
6a-c) and the rotation about a vertical axis (Fig. 6e) matrix that takes into account all successive relative
which are observed to increase in amplitude with rotations (Appendix A). The velocity and acceleration
increased step length. analyses of the two-link chain are developed in terms
of the first and second derivatives computed from the
ABSOLUTE COORDINATES DaERMINATION relative measured joint rotations as shown in Appen-
dix A.
Kinematic analysis
For the purpose of dynamic analysis of gait, the Method of solution
human body can be treated as a chain of rigid links The kinematic analyses of the previous section,
interconnected at joints which allow the links to rotate developed for a general link, can be applied suc-
relative to each other. The system of successive relative cessively to the entire system of twelve links. For this
measurements starts with defining the position of the purpose a program is written in FORTRAN IV that
trunk relative to the chosen fixed coordinate system computes absolute motions of these links from in-
(Fig. 2) and proceeds to express the rotation of each ternally stored harmonic coefficients. To simplify use,
lower limb segment relative to the one proximal to it. the user need supply only six parameters identifying
The computation of the absolute motions takes a the desired gait variability. The first four parameters
similar course; it begins with computing the absolute define the variable. The fifth parameter is a multiple
position of the trunk as a reference member and segment flag designed to save processing time if data of
continues distally to describe the absolute motion of many interconnected segments are needed. The sixth
all segments. defines the step rate.
An analytical method of computing absolute mo- The method of computation in a simplified form is
tions from relative measurements is developed in displayed in Fig. 7. On the basis of the user’s six
Appendix A for a simple chain with two links rotating identifiers, the program interpolates the harmonic
relative to each other about three axes. Figure 12 data of the necessary measured variables to the
depicts such a general linkage consisting of a proximal specified step frequency, generates derivative
link (i - 1). the motion of which is assumed to be coefficients if necessary and reconstitutes the harmonic
known completely and a distal link (i) the motion of data into equivalent time functions. If only measured
which is to be found. The relative rotations are variables are requested, no further processing is done;

Slop Rstor 55 rlopr,min


$p*ad I 110.5 cm0SOC

limo. =- 30
#C

Fig. 8. Stick diagram of the subject walking at 95 steps/min.


Computer generation of human gait kinematics 107

1
Harmonic
Measured: all I
I\ 12
Computed’ 12
,- ,.
\ ,
I \
1 -\ ’
0 ,
#‘_ J \ 8’
i:

~~~~

7-o 50 100
Time. ‘1. of cycle

Fig. 9. Measured and computed forward pelvic accelerations.

HARMONIC ORDER

Computsd Accslcration
Spsctrs
ion % = 223 cm/s*cl

IIAAMONIC ORDEd

Fig. 10. Harmonic spectra of forward pelvic accelerations.


I08 M. Y. ZARRUGH and C. W. RADCLIFIX

i
L
\
-\

en

Time, ,“,f Cycle

Fig. Il. Effects of harmonic content on the reconstitution of ankle tlexion as a time function.

otherwise the program performs kinematic analyses as pares ankle flexioh curves obtained by direct measure
necessary until all requests are satisfied. ment, harmonic synthesis with 20 harmonics and
synthesis with seven harmonics. Some degree of detail
Samples of computed variables is lost when fewer harmonics are used.
Although mass center and joint coordinates of all
segments can be displayed, a “stick” diagram is chosen REFERENCES
for display rather than the coordinates because. it
provides a clear and brief illustration of absolute Ayoub, M. A., Ayoub. M. M. and Ramsey, J. D. (1970) A
stereometric system for measuring human motion. Humcln
motions in the plane of progression. Figure 8 displays
Fucctors12(6), 523-535.
such a diagram of our subject at a step rate of 95 Berstein, MN. A. (1967) The Coordination and Regulation of
steps/min in which the stick representing the foot is Movements. Pergamon Press, Oxford.
deleted for clarity (compare to Fig. 15.4 in Eberhart et Braune. W. and Fischer, 0. (1889) f_&r den den Schwer-
al., 19%). punkt des menschlichen Kihpcrs, mit Riicksicht auf die
Aiistrustung des deutschen Infanteristen. (On the center of
An evaluation of the numerical harmonic gravity of the human body as related to the equipment of
differentiation is made in Fig. 9 in which the experim- the German infantry.) Abh. siichs. Gesellschafi Wissen-
ental and computed (from displacement) horizontal schfren 26,561-672.
pelvic accelerations are displayed as a function of time Cappozxo, A., Leo, T. and Paiotti, A. (1975) A general
computing method for the analysis of human locomotion.
in percent of cycle. Two measured acceleration curves
J. Biomechanics 8, 307-320.
are shown : one includes all harmonics and the other Carlet, M. G. (1872) Essai expgimentale sur la locomotion
the lower twelve harmonics. The corresponding har- humaine. (Experimental essay on human locomotion.)
monic spectra are depicted in Fig. 10 which indicates Annls Sci. nat. 16, l-92.
that the harmonic content is concentrated in the low Cavagna, G. A., Saibene, F. P. and Margaria, R. (1963)
External work in walking. J. appl. Physiol. lS, l-9.
frequency range. This is supported in other studies, Chao, E. Y.-S., Rim, K., Smidt, G. L. and Johnston, R. C.
such as Winter et al. (1974a) and Cappoxzo et al. (1970) The application of 4 x 4 matrix method to the
(1975). At high frequortcies (above the fifteenth har- correction of measurements of hip joint rotations. J.
monic), the spectra resemble those of white noise Biomechanics 3.459-471.
Cheng, I.-S. (1974) Computer-Television Analysis of Biped
which are characterized by uniform magnitude over all
Locomotion. The Communication and Control Systems
frequencies. Because proportionately smaller amounts Laboratory, Department of Electrical Engineering, Ohio
of information and larger amounts of noise are added if State University, Columbus, Ohio.
a large number of harmonics are used, seven to twelve Close, J. R. and Inrnan, V. T. (19S2) The Action of the Ankle
harmonics are considered a good compromise be- Joint. Prosthetic Devices Research Project, Institute of
Engineering Research, University of California, Berkeley.
tween information and noise for most gait variables, Cummings, R. W. and Lamoreux, L. W. (1972) The use of
except ankle flexion which requires at least 20 har- electrical transducers in the measurement of body motions.
monics for adequate reconstruction. Figure 11 com- Bull. Prosthetics Res. 10-18, 46-59.
Computer generation of human gait kinematics 109

Dempster. W. F. (1955) Space requirements of the seated inally published in French by Masson, Paris, 1894.)
owrator. WADC Technical Retort 55-159. Wrizhr Air Milner, M., Dali, D., McConnell, U. S., Brennan. P. K. and
&velopment Center, United Siates Air Force, Gright- Hershler, C. (1973) Angle diagrams in assessment of
Patterson Air Force Base, Ohio. locomotion function. S. Afr. med. J. 47, 951-957.
Drillis, R. (1958) Objective recording and biomechanics of Morris, J. R. W. (19731 Accclerometry - A technique for the
pathological gait. Ann. N. Y. Acad. Sci. 74, 86- 109. measurement of human body movements. J. Biomechanics
Eberhan, H. D., Inman, V. T. and Bresler, B. (1954) The 6, 729-736.
principal elements of human locomotion. In Human Limbs Miiller, E. A. and Hettinger, Th. (1952) Arbeitsphysiologische
and Their Substitutes. (Edited by P. E. Klopsteg and P. D. Unrersuchungen verschiedener Oberschenkei-Kunstbeine.
Wilson et al.), pp. 437-471. McGraw-Hill, New York. (Energy study of different above-knee prostheses.) 2.
Elftman, H. (1967) Basic function of the lower limb. Biomed. Orthop. 81, 525-545.
Engng 2, 342-345. Murray, M. P., Drought, A. V.. Ross, C. and Kory, R. C.
Finley, F. R.. Cody, K. A. and Finizie, R. V. (1969) Loco- (1964) Walking patterns of normal men. 1. Bone Jr Surg.
motion patterns in elderly women. Archs phys. Med. A(2)& 335-360.
Rehabil. 50, 140-146. Murray, M. P. (1967) Gait as a total pattern of movement.
Fischer, 0. ( 1895- 1904) Der Gang des Menschen. Abhand- Am. J. phys. Med. 46.290-333.
lungen der mathematischen Klasse der Kiiniglich- obcrg, K. (1974) Mathematical modeling of human gait : An
Siichischen Gesellschaji der Wfssenchafien. 8 vols. (Vol. 1 co- application of SELSPOT-system. In Biome&nics I C
authored by W. Braune). (Edited by R. C. Nelson and C. A. Morehouse), pp. 79-84.
Gage, H. (1967) Accelerographic analysis of human gait. In University Park Press, Baltimore.
Biomechanics Monograph, pp. 137-152, A. S. M. E., New Pezzack, J. C., Norman, R. W. and Winter, D. A. (1977) An
York. assessment of derivative determining techniques used for
Gerstm, J. W., Orr, W., Sexton, A. W. and Okin. D. (1969) motion analysis. J. Blomechanics 10, 377-382.
External work in level walking. J. appl. Physiol. 26, Pipes, L. A. (1963) Matrix Methodsfor Engineering. Prentice-
286-289. Hall, Englewood Cliffs.
Glanville, A. D. and Kreezer, G. (1937) The characteristics of Ralston, H. J. and Lukin, L. (1969) Energy levels of human
gait of normal male adults. J. exp. Psychol. 21, 277-301. body segments during level walking. Ergonomics 12,39-&i
Gollnick, P. D. and Karpovich, P. V. (1964) Electrogoniomet- Ryker, N. J., Jr. (1952) Glass Walkway Studies of Normal
ric study of locomotion and of some athletic movements. Subjects During Normal Lcwl Walking. Prosthetic Devices
Res. Quart. 35, 357-369. Research Project, Institute of Engineering Reseasch, Uni-
James, U. and &erg, K. (1973) Prosthetic gait pattern in versity of California, Berkeley. Series 11,Issue 20. Berkeley,
unilateral above-knee amputees. Stand. J. rehabil. Med. 5, The Project.
35-M. Smidt,G. L. (1971) Hipmotionand relatedfactorsin walking.
Johnston, R. C. and Smidt, G. L. (1969) Measurement of hip J. Am. phys. Therapy Ass. 51, 9-21.
joint motion during walking: Evaluation of an elec- Smidt, G. L., Deusinger, R. H., Arora, J. and Albright, J. P.
tromyographic method. J. BOG Jt Surg. (A)Sl, 1083-1094. (1977) An automated accelerometry system for p& analy-
Karoovich. P. V. and Wiklow. L. B. (19591 A aoniometric sis. J. Biomechanics 10. 367-375.
s&y of the human foot in standing and ialkiig. U.S.A.F. Suh, C. S. and Radclitie,‘C. W. (1%7a) Synthesis of plane
med. J. 10, 885-903. linkages with use of the displacement matrix. J. Engng Ind.
Kettelkamp, D. V., Johnson, R. J., Smidt,G. L., Chao, E. Y.-S. (B)S9,206-214.
and Walker, M. (1970) An electrogoniometric study of Suh, C. H. and Radcliffe, C. W. (1%7b) Synthesis of spherical
knee motion in normal gait. J. Bone Jt Surg. (A)52, linkages with use of the displacement matrix. J. En& Ind.
775-790. (BP, 215-222.
Kin& F. L., Hillberry, B. M, Hall,A. S.. Jr., Van Sickle, D. C. Sutherland, D. H. and Hagy, J. L. (1972) Measurement of gait
and Harvey, W. M. (1972) Measurement of the total movements from motion picture film. J. Bone Jr Surg.
motion between two body segments--II. Description of (A)S4,787-797.
application. J. Biomechanics 5, 283-293. Townsend, M. A., Izak, M. and Jackson, R. W. (1977) Total
Lamoreux, L. W. (1970) Experimental kinematics of human motion knee goniometry. J. Biomechanics 10, 183-193.
walking. Dissertation Ph.D., Mechanical Engineering. University of California-Berkeley (1947) Fundamental Studies
University of California, Berkeley. of Human Locomotion and Other Material Relating to
Lamoreux, L. W. (1971) Kinematic measurement in the study Design of Artificial Limbs. Prosthetic Devices Research
of human walking. Bull. Prostherics Res. BPR 10-l 5,3-84. Project, Institute of Engineering Research, Final Report to
Lamoreux, L. W. (1974) A self-aligning goniometer for the Committee on Artificial Limbs, National Research
simplified gait analysis. Proc. World Congr. ofISP0 (Int. Council. 2 Vols. Berkeley, The Project.
Sot. Prosthetics and Orthotics), 8-12 Oct. Montreux, Waters, R. L., Morris, J. and Perry, J. (1973) Translational
Switzerland. motion of the head and trunk during normal walking. J.
Lamoreux, L. W. and Todd, F. N. (1974) The importance of Biomechanics 6, 167- 172.
time and distance measurements for analysis of human Winter, D. A., Greenlaw, R. K. and Hobson, D. A. (1972~
walking. Proc. World Congr. UP0 (Int. Sot. Prosthetics Television-Computer analysis of kinematics of human
and Orthotics), 8-12 Oct. Montreux, Switzerland. Gait. Comp. Biomed. Res. 5, 498-504.
Levens, A. S.. Inman, V. T. and Blosser, H. A. (1948) Winter, D. A., Sidwali, H. G. and Hobson, D. A. (1974a)
Transverse rotation of the segments of the lower extremity Measurement and reduction of noise in kinematics of
in locomotion. J. Bone Jr Surg. (A)30, 859-972. locomotion. J. Biomechanics 7, 157-159.
Liberson, W. T., Holmquest, H. J. and Halls, A. (1962) Winter, D. A., Quanbury, A. O., Hobson, D. A., Sidwall. H.
Accelerographic study ofgait. Arch. phps. Med. Rehabil. 43, G., Reimer, G., Trenholm, B. G., Steinke, T. and Shl6sser.
547-554. H. (1974) Kinematics of normal locomotion-A statistical
Liberson, W. R. (1965) Biomechanics of gait: A method of study based on T.V. data. J. Biomechanics 7, 479-486.
study. Arch. phys. Med. Rehabil. 46, 37-48. Wright, D. G., Desai, S. M. and Henderson, W. H. (1964~
Lindholm, L. E. (1974) An optoelectronic instrument for Action of the subtalar and ankle-joint complex during the
remote on-line movement monitoring. In Biomechanics IV stance phase of walking. J..Bone Jr Surg. (A)46, 361-382.
(Edited by R. C. Nelson and C. A. Morehouse), pp. Zarrugh, M. Y., Todd, F. N. and Ralston, H. J. (19741
510-512. University Park Press, Baltimore. Optimization of energy expenditure during level walking.
Marey. E. J. (18951 Mocemenr. Appleton, New York. (Orig- Eur. J. appl. Physiol. 33, 293-306.
1 IO M. Y. ZARRUGHand C. W. RADCLIFFE

APPENDIX A where
Spatial Kinematic Analysis [RI_ I1 = absohtte rotation matrix of member i - 1

Poshion analysis . and


The rotation matrix (Suh and Radcliffe, 1967a and 1967b) b, = initial orientation of measurement axis q.

approach is used to determine absolute displacements from The relative rotation matrix [RJ can now be const+al to
relative rotation measurements. The method is developed for account for measured relative rotation 4. of member i about
a simple two-link chain with a joint a( that allows three u:
successive rotations I$, I& and 4, about nonparalkl axes.
u:v+c U$IV - u,s u,u,v + u,s
This simple chain is represented in Fig. 12. The absolute
motion of proximal link i - 1 is assumed to be completely CRJ= u,u,v+ u,s u:v + c u2u,v- u,s ,
known from previous kinematic analyses which start at the
[ u,u,v- u,s u,u,vf u,s u:v + c 3
rdercncc link having a known absolute motion. The purpose
ofthis analysis is to determine the motion of thedistal link i in W)
terms of the known absolute motion of the proximal link i where
- 1 and the relative rotations in the joint which are measured
about instrument axes 0, v, and w,. c = cos(f#J.)
Having determined the absolute rotation matrix [R,_ II for S= sin(&)
member i - 1. a similar matrix [Rl] can be computed for V= cos(&)
member i by locating the current orientation of the mcasun- uI, IQ and uj = direction cosines a
mcnt axes. The initial orientations of these axes are repre- Because axis v, is distal to nt, its final orientation is found in
sented by II,,,,vu1and wU1and are estimated from photographs terms of rotations of axis u, and member i - 1:
of the subject wearing the measurement apparatus while
standing in relaxed position. Table 3 lists the direction vt= [&- 11C&h. (A3)
&sines of these vectors for all joints relative to the reference A similar rotation matrix to [RJ can be written for relative
coordinate system. The axis II, is rigidly attached to the rotation 4, about axis v, by replacing 4” with 4, in equation
proximal member i - I (Fig. 12). so when this member (A2). The current direction of third measurement axis w, can
execute a rotation [R,_ ,] from the reference position, the now be computed in terms of absolute rotation of member i
vector II,,, rotates with the member to assume its final - 1 and relative rotations about u, and vi:
orientation u,:
wi = ER,- 11CJWWo~ . (A4)
Through the known final configurations of all three measure-
ments axes, the absolute rotation matrix [RJ for link i is
found :
[&I = [JL,I[&lERvlC41. (A51
The matrix [Rn] has the same form as equation (AZ) except
4. is replaced by #,,. Although [R,] is derived for a general
joint having three degrees of freedom, only the hip and ankle
allow as many rotations while the knee allows only two
rotations. Because of small amplitude compared to flexion
angles, some of these rotations, such as ankle inversion, knee
and ankle axial rotations are ignored in theanalysis. With this
simplification, the rotations considered in the study are the
flexion angles of the hip, knee and ankle joints, as well as hip
adduction.
Because point a, is treated as a joint on link i - 1, its
coordinates are determined in the previous stage of analysis
on member i - 1. Therefore, the coordinates of the distal joint
a, + 1and the center of mass c, on link i. are formulated in terms
of all known quantites :
c,= a1+ [&I(co1
- ad 646)
ai+,= ai+ CR&W+ I)- ad. (A7)
where
a, = distal joint on link i - 1

CR,1= absolute rotation matrix of link i


Fig. 12. Two body segments represented as an idealized two- co, = initial position of the center of mass of link i
link chain connected through a point a, by the linkage aO, = initial position of a,
measuring the relative rotations between them. ao,,+ 1, = initial position of a,+ ,.

Table 3. Reference orientation of measurement axes

Direction cosines
Measured
Joint Motion Axis x Y z

Hip Flexion uzo 0 0 1.0000


Hip Adduction v20 l.oooo 0 0
Knee Flexion “20 0 -0.1132 0.9936
Ankle Flexion u40 - 0.3420 0 0.9397
Computer generation of human gait kinematics III

Position analysis should begin at the reference member of from measured displacements. the velocity analysis starts
which absolute motion is defined in terms of absolute with the computation of this members center of mass and
measured angles and coordinates. angular velocities in terms of the computed derivatives.

Velocity analysis Acceleration o&psis


A numerical harmonic differentiation is used to compute This analysis follows the general approach in the previous
the first and second derivatives of measured displacements. section. The absolute angular acceleration of members i’. i”
The velocity analysis is developed in similar fashion to that of and i are expressed as
position. Here the absolute angular velocity Bi of link i is
& = &_, + e,.., x e; + d;., (Al6)
expressed in terms of the known absolute angular velocity
fi,_, of member i- 1 and the derivatives of the relative &=&t&x&+& (A17)
rotations I$,, 9. and I$, measured along the instrument axes. and
Furthermore, the first and second derivatives rotations are
also known through differentiation. For convenience the 0; = & + &’ X Bi + d;,. (Al8)
intermediate links of the measuring device are introduced as Substituting from equation (A8) into (A16) for t?, gives
members i’ and i” (Fig. 12). For example, 4 refers to the
absolute angular velocity of link i’. This velocity is given by 4=8;_, +&-I x &+d;. (At9)
Successive substitutions form equations (Al7) and (A19) into
(A18) eliminates the intermediate quantities and results in an
where
expression for the absolute angular acceleration & in terms of
4-J = absolute angular velocity of number i - 1 measured or computed quantities :
9.s angular velocity of i relative to i - 1 which is = &.u. &=Bi-, +&+6,+diw+8i_, x tj,
Similar expressions are written for member i” and i’ as follows
+ cd,_, + 4.1 x $,
l+y= & + 6, (A9)
+ (&_, + 4. + 4,) x I$,.. (A20)
e, = I$, + 4,. (AtO)
Differentiation of equation (Al2) produces the
The substitution of equations (A8) and (A9) into (AlO) yields acceleration,
8,-B,_, +&+&+&. (All) ? = li x r + 6, x (& x r). (A21)
The velocity of a point c, on member i can be written in Again the vector cross product operations are replaced with
terms of the velocity of a point a, if the position vector r = ci matrix operations as follows :
- a, is differentiated as follows :
f= iC4 + [~Jl?,lh (A221
i = 8, x r. (A12) where
The vector cross product operation t$ x can be replaced with
a matrix product having the following form:
i = [W,]r, (A13)
where

[wi] = [ _i: -;; -i’] 7 (Al4)


and 01,II,, 8, = components of &
The matrix [A] is defined as,
[AI= [atI+ [WiI[ WJ. W3)
Ifthematrixoperationson theright sideofequation (A23)are
and 8,, da, 6, = components of the absolute angular velocity carried out, matrix [A] takes the following form :
8,. The matrix [WJ is called the angular velocity matrix of 4: + e:, e,e, - bi, ri,e, + II,
member i; it is a convenient equivalent to the cross product
operation 8, x (see Pipes, 1963, pp. 163-174). According to [Ai] =/ 8*8* + 8, -&+ei, &t&-U, ,
equation (A13), the velocity of the center of mass L, has the [ e,ci, - 612 e,ti + 0, -ce: + @,I
following form :
(A241
Ci = i* + [WJ(C,i - hi)- (A13

The velocity of the distal joint P,, , is obtained in a similar


Making the use ofequations (A22)and (A23), the acceleration
form to equation (Al5) if c, is replaced by a,+, in this of the center of mass c, can be expressed in terms of the known
equation. Because the velocity of a point on the reference quantities as:
member and its absolute angular velocities are computed ci = iii + [Ai](cor - a& (A25)