A Comparison of Kinematic Recording Instruments
Richard H. Eckhouse,t * M. Ann Penny,2 and Ruth A. Maulucci3
Kinematics, the study of motion, is employed in numerous biomechanics and human performance investigations. The types of instrumentation used in these studies vary at fundamental technical levels, making it difficult to relate results from studies carried out at different laboratories using different instrumentation. A project was designed to compare two commonly used types of kinematic recording techniques, i.e., the 6 df electromagnetic tracker system and the video motion analysis system. A four-level testing and comparison method was conducted involving static and dynamic inanimate objects, as well as human subjects under static and dynamic conditions. It was demonstrated that for rigid body inanimate objects the two systems produce nearly identical values under stationary conditions and are comparable under moving conditions. The systems show only trivial discrepancies in static human body measurements, and perform in qualitatively similar ways on human motion.
KEY WORDS: Kinematics; electromagnetic tracker; video motion analyzer; reaching.
INTRODUCTION
Kinematics, the study of motion, is employed in numerous biomechanics and human performance investigations. The types of instrumentation used in these studies abound, and vary considerably at fundamental technical levels. An early review of recording systems for biomechanical analysis is found in Ayoub.(1) A later comparative analysis of kinematic measurement systems given in 'Tyson and Das.(2) Such a variety of devices makes comparisons among studies conducted at different laboratories with different instrumentation difficult, and limits the usefulness of the results.
An important and much explored area of biomechanics is guided limb motion or reaching for a target in the workspace of the arm. A brief survey of the literature revealed that there are two categories of instrumentation that are very common choices for reaching studies, these being electromagnetic tracking and video motion
'Mathematics and Computer Science Department, University of Massachusetts-Boston, 100 Morrissey Boulevard, Boston, Massachusetts 02125-3393.
2Computerized Biomechanical Analysis, Trabuco Canyon, California.
3M000, Inc., Scituate, Massachusetts.
'To whom correspondence should be addressed.
439
0148-5598i96/12IXL(939$09.50,10 0 1996 Plcnu,n Publishing Corporation
440 Eckhouse, Penny, and Maulucci
analysis. The question arises as to whether the results of studies employing these different types of devices are compatible. Thus, a comparison of the performance of the two types of systems should prove to be of general value.
The first category of systems consists of electromagnetic trackers.(3-7) These provide the 6 df of a moving sensor relative to a fixed source, employing technology that uses low frequency magnetic fields to make these measurements. These systems record angular information directly, with sensors that are attached to the skin surface. There can be slippage of the sensors, particularly in the twist motions, problems can arise due to interference from metal in the environment, and there can be delays due to filtering.
The second category, video motion analyzers, is made up of two- and threedimensional human performance systems.(8'9) Body segment points are located with retroreflective spherical markers attached by double-sided tape, and the systems automatically track these markers through successive frames, assuming adequate background contrast. Each camera sees a marker in two dimensions; a marker must be in the field of view of at least two cameras simultaneously in order for its threedimensional location to be calculated. In some systems, if a marker becomes occluded, the system automatically supplies the missing point by interpolation. Typically, system resolution and accuracy are high. Software for data capture, threedimensional calibration, computation of the three-dimensional coordinates of the target markers, interactive graphics for plots and stick figures, statistical routines, and signal processing is often available, and data reduction is rapid. Some of these systems demand a controlled environment, since the system requires multiple electrical outlets, a room with a matte finish, heavily draped windows, and fluorescent light. An optional feature in some of the systems is that it is possible to record without markers and then manually digitize each frame; the trade-off, however, is that the operator must locate joint centers. As a compromise, some systems allow points of interest on the subject to be tracked semi-automatically without markers, requiring that the operator initially identify the desired points; the system then predicts the location of these points in subsequent frames.
There are many examples of reaching studies that used electromagnetic or video systems. The use of electromagnetic monitoring to quantify eye-head-hand coordination during reaching tasks has been reported.('�) Free(11) and constrained(12) arm and leg reaching movements were examined using a four-receiver electromagnetic device. A computerized system centered around a 6 df electromagnetic device was build by Eckhouse et al.(13) to facilitate reaching in preschool children with cerebral palsy by modifying the trajectory of the arm. A magnetic kinesiograph was used to compare slow arm movement trajectories in patients ' with Parkinson's disease to those of normal subjects.(14) An electromagnetic transmitter was used to examine whether the straight line paths of two-joint planar reaching movements reflect constraints associated with perception or movement production.(15) High-speed film was digitized to discuss the effects of load and variation on the autoregressive integrated moving average model for muscle signals.(16) Three high-speed cameras were used to identify the curvature-speed relationship in the reaching movements of infants.07)
Instrument Comparison 441
A project was designed to compare these two commonly used types of kinematic recording techniques, i.e., the 6 df electromagnetic device and the video motion analysis device, in guided limb motion tasks. The specific member of the former group that was selected was the Ascension Flock of Birds (FOB), and the Ariel Performance Analysis System (APAS) was chosen from the latter group. The goal was to determine how close these two very different types of devices operate, and whether the results obtained from one system could be compared with the results from the other. There are several possible sources of measurement errors or differences for the two systems, particularly in applications involving human subjects. They could be caused by system setup, sampling rate, resolution, accuracy, receiver and marker attachment to the body, and receivers and markers moving with the skin surface. Digitizing errors and magnetic field corruption contribute additional complicating factors. To isolate the possible different causes, a four-level testing and comparison method was conducted involving static and dynamic inanimate objects, as well as human subjects under static and dynamic conditions.
MATERIALS AND METHODS
Throughout this discussion, trajectory refers to a time series of points that define the position of a translational coordinate or angle throughout its interval of motion. A functional reach is a task oriented limb movement, such as reaching for a target. With regard to human subjects, x is the medial/lateral, y is the anterior/posterior, and z is the superior/inferior direction.
Kinematic Systems
The FOB is a real-time, electromagnetic mechanism that provides the location (x, y, z) and orientation (a-azimuth, e-elevation, r-roll), i.e., 6 df, of moving receivers relative to a fixed transmitter. Output is in the form of time series signals that uniquely determine the location and orientation of the receivers relative to the transmitter. The receiver is a small cube of about 2.5 cm that weighs less than 0.03 kg. The transmitter assembly is a larger cube about 9.5 cm weighing about 1.8 kg. The system provides hemispherical measurement within a 1 m radius. System resolution is 0.1 deg rotational and 0.08 cm translational; accuracy is approximately 0.5 deg for rotations and 0.25 cm for translations. Sampling rate is 100 Hz per receiver. The output is connected via a high speed asynchronous serial RS-485 communications adapter to a microcomputer for immediate processing. Up to four FOB receivers were used simultaneously in this project, each receiver attached to an individual object or body segment, allowing the recording of the three translational and the three rotational coordinates of that object or segment. Thus, the exact . location and orientation of the object or segment was specified throughout the interval of motion.
The APAS is a three-dimensional video motion analysis system. One or multiple cameras can be used, markers are optional, and automatic or manual digitizing is supported. The image sequences are stored, in picture form, on a computer for
442 Eckhouse, Penny, and Manlucci
subsequent processing. The system permits frame grabbing, frame retrieval, data compression, smoothing and filtering options, and zooming. Viewing and graphing modules for the digitized activity, measured or computed, are available. Two variable shutter cameras were used in this project, each operating at 60 frames per sec. Several types of reflective markers were used to assist in identifying the points of interest: a finger cot with reflective tape, large and small reflective balls, and a splint with a small reflective ball attached to each end. Selected points were digitized in each frame, producing the r, y, and z translation trajectories, i.e., the location of each coordinate over time, of the points.
A method was devised so that the FOB and APAS could be synchronized and used concurrently. Each trial proceeded as follows. A large light emitting diode (LED) display was placed in the viewing field of the two APAS cameras. The two cameras were started. The trial number was presented to the cameras manually by placing a clipboard containing the written number in front of the cameras for a few seconds. By means of a keystroke, three events occurred simultaneously, namely, a tone was emitted, the LED display changed, and FOB data collection was initiated. The data were collected for 6 sec. At that instance, the LED display changed again and the FOB stopped automatically. The cameras were turned off manually. Thus, the two APAS frames in which the LED changed corresponded to the initial and final data points of the FOB. This allowed the signal points of one
system to be matched in time with the signal points of the other system, and in addition synchronized the two APAS cameras.
Four-Level Testing Protocol
A formal protocol was developed to permit comparison of the two systems. System performance was examined at four levels, each increasing the difficulty of obtaining system compatibility. The first level, termed static inanimate, was to determine the , y, and z translational distances between two fixed points, the fixed flexion angle formed by a goniometer, and the fixed twist angle displacement made by a goniometer. In the second level, dynamic inanimate, trajectories were obtained for the three translation signals of a randomly moving point, the randomly changing flexion angle formed by two rods, and the randomly changing twist angle produced by a rod. For the third level, static animate, a human subject held his or her arm in various fixed positions intermediate to a functional reach; the three translational distances between selected body landmarks, joint flexion angles, and angles formed by body segments and the coordinate axes were obtained. At the fourth level, dynamic animate, the human subject made simple functional reaches; the x, y, and z translation trajectories of selected body landmarks, joint flexion and twist angle trajectories, and trajectories.of angles formed by body segments and the coordinate axes were acquired.
The levels and sublevels of the protocol are abbreviated as follows: I. Static Inanimate
A. Translation
B. Flexion Angle
C. Twist Angle
Instrument Comparison 443
II. Dynamic Inanimate
A. Translation
B. Flexion Angle
C. Twist Angle
III. Static Animate
IV Dynamic Animate
Levels III and IV were conducted with human subjects. No attempt was made to control for age, gender, handedness, stature, or weight, since the objective was simply to compare data simultaneously acquired by the FOB and the APAS. Informed consent was obtained from each subject prior to the start of any testing.
Level LA
Level LA was to use the FOB and the APAS to calculate the x, y, and z translational distances between two fixed points. An APAS large reflective ball was attached to the top of each of two FOB receivers, and the two marker/receiver combinations were set at oblique x, y, and z distances. The x, y, and z distances were measured three times in short succession with a steel measuring tape, calculated for each sample acquired by the FOB during the 6-sec data collection interval, and calculated for each frame acquired by the APAS during the six-second data collection interval. Three trials were made, with the marker/receiver combinations placed in different locations each time.
Level LB
Level LB was to calculate the fixed flexion angle formed by a plastic full-circle universal goniometer. An APAS large reflective ball was attached to the top of each of two FOB receivers. One of these was placed at each arm end of a goniometer, and one APAS large reflective ball was placed at the vertex of the goniometer. The goniometer was set to an arbitrary angle. The goniometer angle was read three times, and was calculated for each FOB sample and each APAS frame during the 6-sec collection interval. Fourteen trials were made, with the goniometer at different angles and in different states of rotation and tilt relative to the APAS cameras.
Level LC
Level I.C was to calculate the fixed twist angle displacement made by a goniometer. An FOB receiver was attached to the center of the flat side of an APAS splint that had a small reflective ball attached to each end, with the receiver wire perpendicular to the splint. The-edge side of the splint was placed along one arm of a goniometer, with the center of the edge at the vertex of the goniometer. Note' that in this configuration, if the splint combination were attached to a segment perpendicular to the goniometer at its vertex, the flexion angle of the goniometer arm would be equivalent to the twist motion of the segment. During the 6-sec data collection interval, the goniometer was set in succession to two different arbitrarily
444 Eckhouse, Penny, and Maulucci
selected angles. The two angles were read three times each, and calculated for each FOB sample and APAS frame in which they occurred, i.e., for approximately 3 sec each. The means of the two angle values were used to calculate the mean twist angle displacement. Eight trials were made, with the goniometer skewed differently with respect to the APAS cameras.
Level HA
Level II.A was to obtain the trajectories for the x, y, and z translation signals of a randomly moving point. An APAS large reflective ball was attached to the top of an FOB receiver, and the marker/receiver combination attached to the end of a long rod. The rod was manually translated randomly in three dimensions during the data collection interval. The x, y, and z trajectories of the marker/receiver object were calculated from the FOB and APAS data, and then graphed. Four trials were made, with different random motions and speeds of the rod.
Level II.B
Level II.B was to obtain the trajectory of a randomly changing flexion angle formed by two rods. An APAS large reflective marker was attached to the top of each of two FOB receivers. One marker/receiver combination was put at each end of two joined rods, and an APAS large reflective marker was put at the vertex of the joined rods. During the data collection interval, the flexion angle of the rods was changed while the rods were translated, all with continuous random motion made manually. The flexion angle trajectory of the rods was calculated from the FOB and APAS data, and graphed. Three trials were made, with different translating and angle motions of the rods.
Level II. C
Level II.C was to obtain the trajectory of a randomly changing twist angle of a rod. An FOB receiver was attached to the center of an APAS splint with a small reflective ball at each end. The splint was put on the rod as a crosspiece. During the data collection interval, the twist angle of the rod was changed while the rod was translated, all manually with continuous random motion. The twist angle trajectory of the rod was calculated from the FOB and APAS data, and graphed. Three trials were made, with different translation and angular motions of the rod.
Level III
Level III consisted of a human subject holding his or her arm at the initial, intermediate, and final positions of a simple three-dimensional functional reach. Targets of various shapes and in various locations within the workspace of the arm
Instrument Comparison 445
from a seated position were used. All reaches were made at natural speeds, as established by the subject.
The basic reach consisted of a decision to access a target in the workspace, followed by visually locating the target, and then using visually guided motion to bring the arm naturally to a successful access of the target. During the reach, the torso remained stationary and against the chair. The proximal end of the clavicle was fixed, meaning that it did not translate. Clavicle, shoulder, elbow, and wrist motion was allowed. Accuracy was important, reaction time was not, and movement time was whatever was natural for the individual. The initial position was with the upper arm perpendicular to the ground, the elbow at 90 degrees, and the forearm midway between pronation and supination. The final position was with the pad of the middle finger on the target and the palm facing away from or toward the subject. An intermediate position was any position between the initial and final positions that occurred naturally during the reach. The forward distance of all targets from the shoulder was the same and was determined as follows. It was the distance at which the shoulder had to be placed from a point that was 16 cm above and laterally aligned with the shoulder to induce full arm extension to touch the point.
The subject was instrumented with four FOB receivers, one on each segment of the arm, i.e., the hand, forearm, upper arm, and clavicle. The receivers were numbered distally to proximally 1, 2, 3, and 4. It was found that the best results were obtained by first mounting the receiver on a tongue depressor, then securing it to the body segment with medical tape, and then placing a light net stretch-andhold first-aid sleeve over it. The subject was simultaneously instrumented with several APAS reflective markers, i.e., a finger cot, forearm splint, large elbow ball, upper arm splint, large shoulder ball, and large clavicle ball. The large elbow, shoulder, and clavicle balls were not placed on the joints, but rather served as a cue for locating the joint centers during later digitizing. The forearm and upper arm FOB receivers and the corresponding forearm and upper arm APAS splints were placed at the same point longitudinally on the limb segment so as to minimize actual differences along the segment of the forearm pronation/supination and shoulder medial/lateral rotation twist angles. Figure 1 illustrates the subject instrumentation, complete with receivers and markers. Six subjects were used. Three targets in different locations and requiring different final positions were used. For each target, the subject assumed the three positions for six seconds each.
The following calculations were made from both the FOB and APAS data, using all samples or all frames of the 6-sec data collection interval. The , y, and z translational distances between each adjacent pair of FOB receivers was calculated. Wrist flexion/extension, wrist radial/ulnar deviation, and elbow flexion/extension angles were also calculated. Finally, the upper arm X-axis angle, upper arm Y-axis angle, and upper arm Z-axis angle were calculated, these being defined relative to an inertial moving coordinate system, with origin that was attached to and moved with the shoulder at the acromion and axes that remained directed as and parallel to those of the FOB transmitter.
446 Eckhouse, Penny, and Maulucci
Fig. 1. Instrumentation of the subject with the FOB receivers and the APAS reflective markers.
Level IV
Level IV consisted of a human subject making a simple three-dimensional functional reach. The basic reach, targets, and instrumentation were as defined for Level III. The reach was made during the 6-sec data collection interval. During this interval, the x, y, and z translation trajectories of the end effector, wrist, elbow, and shoulder were obtained. Trajectories for the wrist flexion/extension, wrist radial/ulnar deviation, forearm pronation/supination, elbow flexion/extension, upper arm Xaxis, upper arm Y-axis, upper arm Z-axis, and shoulder medial/lateral rotation angles were also obtained during the 6-sec interval. Graphs for all trajectories during the actual movement time, i.e., the time from the initiation of arm movement to the touch of the target, were produced. Eighteen targets, differing in location or final position required, were used.
RESULTS
Several types of analyses were performed on the data acquired with the manual system, the FOB, and the APAS in the four-level testing. It was necessary to limit the number of trials for incorporation into the analyses because the APAS data were manually digitized which is an extremely time-consuming process for this type of application. Representative trials were randomly selected from each level or
sublevel.
Table I gives measurement values obtained from Level I testing for the measuring tape or goniometer (referred to as the manual system), the FOB, and the APAS. The a y, and z values (centimeters) are for one arbitrarily selected trial of Level I.A. They are calculated by taking the average of the three tape measure-
Instrument Comparison
447
Table I. Measurement Values of the Translational Distances
Between Two Fixed Points (Centimeters), a Fixed Flexion
Angle (Degrees), and a Fixed Twist Angle Displacement (Degrees) for the Manual System, the FOB, and the APAS
Manual FOB APAS
System
x 37.4 37.7 38.1
y 55.9 56.1 55.6
z 20.0 21.0 20.5
Flexion angle 135 135 136
Twist angle 46 47 47
ments, the average of the 600 FOB samples, or the average of the 360 APAS frames. The flexion angle values (degrees) are for a trial in Level I.B in which the goniometer was perpendicular to the ground, but at a 45 degree angle outward with respect to the APAS cameras. They are calculated by taking the average of the three goniometer readings, all of the FOB samples, or all of the APAS frames. The twist angle values (degrees) are calculations of the displacement angle for a trial in Level I.C in which the goniometer was rotated and tilted relative to the plane of the APAS cameras. They are calculated from the average of the three goniometer readings, all of the FOB samples, or all of the APAS frames for each of the two composite angles. A one way analysis of variance applied to the five values obtained with each of the three techniques revealed no differences in the means among the manual system, the FOB, and the APAS (p > 0.05).
Figures 2, 3, and 4 afford a visual inspection of results of the Level II testing. Trajectory graphs of the x, y, and z translation signals for one trial of Level II.A for the FOB and the APAS, respectively, are given in Fig. 2a,b. One trial of flexion angle trajectory graphs from Level II.B is given in Fig. 3a,b, respectively, for the FOB and APAS. Graphs of the twist angle trajectory at Level II.C for one trial are given in Fig. 4a,b for the FOB and APAS, respectively.
Table II contains the measurement values (centimeters for translations and degrees for angles) obtained from Level III testing for the FOB and the APAS for the initial, intermediate, and final positions of one reach trial. Each value is the average of the 600 FOB samples or the 360 APAS frames. The target in this trial was located 8 cm above and 8 cm to the left of the shoulder. The final position for this trial was with the palm facing away from the subject. A one way analysis of variance applied to all of the translation values obtained with each of the two techniques revealed no significant differences (p > 0.05) between the FOB and the APAS. Similarly, no significant differences (p > 0.05) between the FOB and APAS were observed for the angle values.
Figures 5 and 6 show Level IV testing results for the 12 translation trajectories and the eight angle trajectories. Graphs for one trial from one subject are shown in Fig. 5a-d for the FOB translations and angles and the APAS translations and angles, respectively. The target in this case was located at shoulder level and 16 cm to the right of the shoulder, with a final position of the palm facing toward the subject. Figure 6a,b, respectively, for translations and angles, shows this same trial as the FOB values plotted against the APAS values during the movement time. Specifically, for each point during the movement time of the reach, each of the 12
APAS translation values is paired with" the corresponding FOB translation value and all pairs plotted on the same graph; similarly, for each sample, the eight APAS and corresponding FOB angles are paired and plotted on a second graph. These graphs offer a further examination of the similarity of the two systems, in that the closer the data points in the graphs can be fitted by a straight line with slope 1, the more alike are the values of the two systems.
Fig. 3. Trajectory of a randomly changing flexion angle from (a) the FOB and (b) the APAS.
Table III is a chart of a statistical analysis of the Level IV results over all six subjects for the target just defined. Six subsets of the set of 20 translation and angle trajectories were selected, these being the end effector translation; elbow translation; all four joint translations taken together; elbow flexion/extension angle; upper arm X, Y, and Z axis angles; and all eight angles taken together. For each
Fig. 4. Trajectory of a randomly changing twist angle from (a) the FOB and (b) the APAS.
subset, the signed difference between corresponding FOB and APAS values for each sample during the movement time of the reach trial for every subject was calculated. Descriptive statistics of mean, median, and standard deviation were applied to these differences, for each subset. A mean close to 0 implies that there is no bias between the two systems in that it indicates that neither system is consistently operating at a higher or lower range than the other, although it does not necessarily signify that
Instrument Comparison
451
Table II. Measurement Values of the Translational Distances (Centimeters) and Angles (Degrees) at
Three Positions of a Functional Reach for the FOB and the APAS.
Initial Position Intermediate Position Final Position
FOB APAS FOB APAS FOB APAS
Receiver 1-2 x 0.4 0.1 -2.4 -0.7 -4.1 -1.9
Receiver 1-2 y 11.4 8.8 8.8 9.5 9.7 9.7
Receiver 1-2 z -0.3 -0.1 -6.8 -7.4 -4.6 -5.4
Receiver 2-3 x 2.1 0.2 -5.2 -11.4 -8.5 -7.2
Receiver 2-3 y 24.8 25.5 30.2 28.2 37.2 36.2
Receiver 2-3 z 15.7 16.3 -7.8 -7.3 -15.0 -14.4
Receiver 3-4 x -18.5 -18.5 -19.0 -18.4 -18.4 -21.4
Receiver 3-4 y -8.2 -7.6 4.1 -1.7 9.0 4.6
Receiver 3-4 z 15.1 11.3 10.2 10.3 1.1 3.9
Wrist flexion/extension 5 0 -9 6 -12 10
Wrist radial/ulnar 5 2 6 10 2 11
Elbow flexion/extension 81 85 83 87 129 135
Upper arm X-axis 91 88 82 81 80 75
Upper arm Y-axis 83 89 128 135 161 163
Upper arm Z-axis 173 178 140 134 106 96
the differences between the system values are small. Finally, a reliability coefficient(18,19) was calculated for each subset, using the actual measurement values given by the two systems. This coefficient, ranging from 0.0 to 1.0, is a measure of the capability of the FOB and APAS to produce the same value when operating under uniform conditions, with 1.0 indicating perfect reliability.
DISCUSSION
The FOB and the APAS have been compared under different conditions of operation. In Level I, it was possible to evaluate these two instruments against traditional devices used in conventional anthropometry, namely, the steel measuring tape and the plastic goniometer, thus allowing the results to be considered with respect to commonly accepted actual values. In general, however, since there are no standard kinematic instruments with which to validate the results, this study investigates how close the performances of the FOB and APAS are to each other, rather than how close either is to the true physical measurement.
Level I, the easiest level in which to obtain system compatibility since it deals with rigid bodies and no motion, shows the FOB and APAS to be operating close to each other and close to the actual physical measurements. Level II becomes more difficult with the introduction of motion. However, the systems are still operating with visually comparable results for both translation and angle trajectories.
Level III introduces to the systems the complexities in dealing with the human body. Nonetheless, the FOB and the APAS still show no significant incompatibilities' in this static case. This is particularly impressive since due to the inherent operation of the two systems, they do not always calculate from the same quantities when used with a human subject. For example, a tilt in the receiver caused by the slants and contours of the body segment to which it is attached will influence the angle
Fig. 5. Functional reach trajectories for (a) FOB joint translations, (b) FOB joint angles, APASjoint translations, and (d) APAS joint angles.
Instrument Comparison
455
Table III. Statistical Analysis of the FOB and APAS Differences (Centimeters for Transla
tions and Degrees for Angles) for Functional Reaches From Six Subjects
Mean Median SD Reliability
End effector translation 1.22 0.88 1.94 0.997
Elbow translation -0.07 -0.03 1.48 0.999
All joint translations 0.12 0.20 2.12 0.998
Elbow flexion/extension angle 4.17 6.11 5.82 0.954
Upper arm X-, Y-, Z-axis angles 2.67 2.47 4.28 0.991
All joint angles 4.33 3.29 15.49 0.980
measured by the FOB. Furthermore, the anatomical angles do not correspond to the Euler angles that are directly collected by the FOB, and must be calculated, often from more than one receiver; slight misalignments can impact on these calculations. With regard to the APAS, misses in the location of the joint center on the limb segments, with either manual or automatic digitizing, will influence the angle measurements. Finally, for the FOB, the joint angles are calculated from angle values, whereas for the APAS they are calculated from positional values of the vertex and sides of the angle; thus, numerical discrepancies might be expected to arise from the mathematical calculations.
At Level IV, the full array of complicating factors inherent in the human body are present. Issues such as skin rotations, muscle activity, and tissue masses all have the potential to contribute errors in the acquired data and challenges to system compatibility. From the graphs, it can be observed that the shapes and displacements of corresponding signals are quite similar. The reliability coefficients indicate that overall the systems produce close results for both translation and angle trajectories.
It is important to note that the data were taken from the systems at the most rudimentary level of each system, prior to the application of any system analysis software, and then processed using the same analysis software in both cases. In particular, the FOB data were taken as the x, y, z locations and the a, e, r angles of the receivers, and the APAS data were taken as the x, y, and z locations of the end effector, wrist, elbow, shoulder, and sternoclavicular joint, and of the two reflective balls on the forearm and the upper arm crosspieces. Both sets of system data were then analyzed by the same software to derive the. y, and z translation trajectories of the end effector, wrist, elbow, and shoulder, and trajectories for the wrist flexion/extension, wrist radial/ulnar deviation, forearm pronation/supination, elbow flexion/extension, upper arm X-axis, upper arm Y-axis, upper arm Z-axis, and shoulder medial/lateral rotation angles. As explained above, the nature of data required different calculations on the FOB and APAS data to attain the same measurement. With this exception, however, any differences are due to the basic operation of the systems, and not to discrepancies in signal processing algorithms or routines.
CONCLUSION
It may be concluded that the electromagnetic tracker and video motion analysis systems perform closely enough on static inanimate objects to allow quantitative data to be compared, and permit dynamic inanimate objects to be compared quali-
456 Eckhouse, Penny, and Maulucd
tatively. The systems present no problems for the static animate case, and can be used interchangeably. Qualitative comparisons can be made in the dynamic animate case, and for studies interested in basic trajectory shapes and displacements the two systems are demonstrably comparable.
ACKNOWLEDGMENTS
The authors wish to thank Dr. Gideon Ariel for his extensive technical assistance in adapting the Ariel Performance Analysis System for this study. This work was supported by NASA contracts NAS9-18514 and NAS9-18915.
REFERENCES
1. Ayoub, M.M., Human movement recording for biomechanical analysis. Iraa J. Prod. Res. 10(1):35-51, 1972.
2. Tyson, J.N., and Das, B., A comparative analysis of anthropometric and kinematic measurement systems. In (B. Das, ed.), Advances in Industrial Ergonomics and Safety II, Taylor & Francis, pp. 301-308, 1990.
3. Logan, S.E., Vannier, M.W., Bresina, S.J., and Weeks, P.M., Wrist kinematic analysis using a 6 degree of freedom digitizer. Proceedings of the Seventh Annual Conference of the IEEE Engineering in Medicine and Biology Society, 13-17, 1985.
4. Bresina, S.J., Vannier, M.W., Logan, S.E., and Weeks, P.M., Three-dimensional wrist imaging:
Evaluation of functional and pathologic anatomy by computer. Clin. Plast. Surg. 13(3):389-405, 1986. 5. An, K.-N., Jacobsen, M.C., Berglund, L.J., and Chao, E.Y.S., Application of a magnetic tracking
device to kinesiologic studies. J. Biomech. 21(7):613-620, 1988.
6. Logan, S.E, Groszewski, P., Krieg, J.C., and Vannier, M., Upper extremity kinematics assessment
using four coupled six degree of freedom sensors. ISA Proceedings 75-81, Paper #88-0211, 1988. 7. Pearcy, M.J., and Hindle, R.J., New method for the non-invasive three dimensional measurement
of human back movement. Clin. Biomech. 4:73-79, 1989.
8. Bravo, 0., Angular levers and power in the human body. IEEE Eng. Med. Biol. 11:55-60, 1992.
9. Diener, H.-C., Bacher, M., Guschlbauer, B., Thomas, C., and Dichgans, J., The coordination of
posture and voluntary movement in patients with hemiparesis. J. Neural 240:161-167, 1993.
10. Bak, D.J., Electromagnetic monitoring quantifies motion behavior. Design News 88-91, 1986.
11. Maulucci, R.A., Optimal Workspace Design. Technical Report, National Aeronautics and Space
Administration, Contract NAS9-18514, Lyndon B. Johnson Space Center, Houston, TX, 1993.
12. Maulucci, R.A., Nonstandard Functional Limb Trajectories. Technical Report, National Aeronautics
and Space Administration, Contract NAS9-18915, Lyndon B. Johnson Space Center, Houston, TX,
1995.
13. Eckhouse, R.H., Leonard, E.L., Zhuang, Q., and Maulucci, R.A., Improving reaching in preschool children with cerebral palsy through regulated feedback. IEEE Transact. Rehab. Eng. 2(3):147-157, 1994.
14. Isenberg, C., and Conrad, B., Kinematic properties of slow arm movements in Parkinson's disease. J. Neu of 241:323-330, 1994.
15. Flanagan, J.R., and Rao, A.K., Trajectory adaptation to a nonlinear visuomotor transformation:
evidence of motion planning in visually perceived space. J. Neurophysiol 74(5):2174-2178, 1995. 16. Sherif, M.H., Gregor, R.J., and Lyman, J., Effects of load on myoelectric signals: the ARIMA rep
resentation. IEEE Transact. Biomed Eng. 28(5):411-416, 1981.
17. Fetters, L., and Todd, J., Quantitative assessment of infant reaching movements. J. Motor Behav. 19(2):147-166, 1987.
18. Fleiss, J.L., The Design and Analysis of Clinical Experiments, John Wiley and Sons, New York, 1986. 19. Olkin, I., and Pratt, J.W., Unbiased estimation of certain correlation coefficients. Amer. Math- Stat. 29:209-211, 1958.