2. INTRODUCTION
While a large amount of literature exists regarding the physiological
issues of HMD systems with tracking, little empirical work has been
published regarding systems where head-tracking is not performed.
This seems to suggest that the advantages supplied by the HMD indicate
that head-tracking must accompany them. However, no correlation has
been derived to support this notion. This report is intended to
provide a first-pass set of empirical evidence on a tracking task
using a telerobotic vision system displayed on either an HMD or a
panel-mounted visual display. From the study's data, an understanding
of the limitations and benefits of display method is reviewed.
The use of Head-Mounted Displays (HMDs) has been commonly designed
into man-machine systems in which head-tracking contributes to the
task of interest. Is this the only benefit of using HMDs? This study
removes the tracking aspect from the HMD in order to isolate
differences of display type (HMD vs. panel display) on user
performance in a tracking task.
The apparatus used in this experiment consists of a joystick as the
sole means of input control while comparing the performance in a
tracking task displayed on a head-mounted display (HMD) and a static
console display. Aspects of each device have been previously studied
to assess their worth in providing human interaction with computers.
This section individually analyzes the output devices (HMD and fixed
display) and the input device (spring-loaded joystick).
Advantages of HMDs
There are potentially many advantages of using HMDs over static
displays. Military aircraft have been using HMDs for over 20 years to
assist fighter pilots in maintaining visual awareness outside the
cockpit and facilitating quick decision-making in air-to-air combat.
HMDs may "display information directly in the line of sight of the
pilot, focused at infinity and overlaid on the outside world, such
that the pilot can acquire and interpret the information without
constant shifting of visual accommodation and attention from the
outside world to the instruments below the glare shield and back
again" [Taylor 1992]. Commercial aviation HUDs will eventually be
made smaller and moved to either a helmet or a pair of specially
designed lightweight glasses connected via a fiber optic to a symbol
generator.
HMD vs. Static Display
Arthur and Booth [1993] conducted experiments comparing HMDs with
tracking and fixed displays with head tracking in a tree tracing task.
After gaining experience viewing 3D graphical objects on both displays
with the head tracking capability to change their perspective, data
from a survey was collected before conducting the experiment. From
the survey, the majority of the users preferred head coupling without
stereoscopic viewing over stereoscopic viewing without head coupling.
It was also noted that the head coupling apparatus was awkward for the
subjects to use, and that the actual task would dictate its usability.
Although the survey placed little importance on stereoscopic viewing,
the experiment results showed that head coupling with stereo was the
fastest, while head coupling without stereo was significantly slow.
As an addendum to their experiment, the effects of lag and frame rate
on performance were considered only for the case of head coupling with
stereoscopic viewing. Lag is due to processing the motionary input
and rendering the output; network lag did not need to be considered,
since the workstation was isolated from any network. The results
suggest that for similar tree tracing tasks, lag above 210
milliseconds will produce a decline in performance. In our study, lag
does not become a distraction since the telerobotic visual system does
not suffer from network traffic.
Perspective Views
Our study only concerns itself with tracking in two-dimensional space.
Typical HMD applications tend to display three-dimensions. If this
extra dimension were to be considered in our tracking task, a number
of additional factors are exemplified by the following studies.
In an aircraft landing study, binocular cues, such as convergence and
stereopsis, were found to only be effective in the last few seconds of
landing the aircraft; monocular cues seem to be the important ones for
visual space perception during approach to landing [Dorfel 1982]. In
addition, near-visual response, particularly lens accommodation,
results in a minified retinal image leading to the runway appearing
more distant and a perception of undershooting [Randle, Roscoe &
Petitt 1980]. The most important cues for pilots during final
approach and landing are retinal image size (reflected in distance and
height judgment), shape of the runway (which affects slope angle
judgment), and motion parallax (which gives the time history of these
factors). Results so far show that pilot subjects were able to judge
distance, height and slope angle from computer generated landing
approach scenes with remarkably good accuracy [Dorfel 1982]. Studies
have shown that, at large viewing distance, size constancy is the most
accurate method of depth perception, with movement parallax becoming
preferably at medium range, and the binocular cues taking over at
fairly close range [Buffett 1986].
The most prevalent problem normally associated with flat-panel image
displays is the two-dimensional viewing of a three-dimensional object.
Textures on objects have typically been used to establish depth and to
provide anchors. This has mainly been helpful in colorized imagery,
but not as effective in gray-scale imagery [Dorfel 1982].
Field of View
Alfano and Michel [1990] reported that restricting the normal field of
view leads to perceptual and visuomotor decrements. A series of
eye-hand coordination tasks were performed using goggles that
restricted the field of view to 9, 14, 22 and 60 degrees. Although
Pelli [1986] proposed that a 22 degree restriction would cause no
measurable performance decrements, Alfano and Michel confirmed Dolezal
[1982] reports that performance suffers even at these field of view
restrictions. While the 60 degree field of view restriction yielded
significantly better performance than the others, all of the degree
restrictions chosen caused a sense of disorientation in the subjects'
depth and size judgment. Studies conducted by the Naval Training
Equipment Center in the early 1980's conclude that the maximum
single-channel field-of-view acceptable for flight applications is
about 90 degrees when displayed on a virtual or flat screen display
[Chambers 1982]. Increasing the amount of peripheral information (by
increasing the field of view) allows the subject to construct an
overlapping sequence of fixations in memory, which aids visuomotor
performance.
Area of Interest
Area of interest processing is the computational method of enhancing
an image in the area that a viewer is looking [Chambers 1982].
Developed for field-of-fire trainers by military contractors, and used
effectively in the British Army Training Command, this system would
allow the viewer to see the detail required near the target, while
significantly less computing was performed on the scene out of the
central field-of-view, referred to as the area of interest [Chambers
1982]. The area of interest is approximately 60 degrees, and within
it is the foveal focus field-of-view, approximately 2 degrees of arc.
Reduction in the scene content (number of polygons) outside the area
of interest reduces the amount of computer power required while
providing maximum fidelity to the scene where needed. When the viewer
is looking in another direction eye tracking devices would drive the
computer image generator to enhance the area of interest on the
display where the viewer is looking while reducing the content of the
non-viewed area of the display. This reduces computational time and
the excess would be used to preprocess information for the
field-of-view being enhanced.
The area of interest in our experiment is the tracking cursor that is
always positioned in the center of the display. No other objects on
the foreground of the path distract the subjects from this area of
interest, and the cursor is always clearly in view.
Peripheral Information
But, displays are limited in representing a wide field of view. If
the actual space is scaled proportionally to the monitor's display
space, the display resolution may sacrifice detail. However, if only
a partial view of the actual space is represented in the display,
tracking tasks may submit to the same problems of small field of view.
However, Flach, Hagen and O'Brien [1990] reported that the
proportional mapping of actual space to display space was a safer
choice in tracking tasks than non-linear mappings. In their
experiment, comparisons of performance in a positioning task were made
between three different mappings of visual display to movement space.
In the normal display condition, displayed distance between targets
was proportional to the actual distance. In the split screen
condition, 66.5% of the initial distance to the target was mapped to
half of the visual space and the remaining 33.5% of the distance
(containing the target) was mapped to the other half of the visual
space. Finally, in the logarithmic condition, there was a logarithmic
mapping from actual to visual space. Subjects were instructed to
position a cursor to a target using only horizontal movements. The
width of the target varied in three defined sizes. The results
indicate that performance is not significantly different for the three
display mappings when pursuing the largest target. As the size of the
target decreased, linear mappings tended to yield better performance.
Color
Eye fatigue has been a major problem in using HMDs and can partially
be attributed to poor color choices. Cyan, white, and green have been
proposed as the best colors to use in a display. These colors
provided easily distinguishable sharp images. Magenta was found to be
the most uncomfortable color, being somewhat harsh, and at first sight
"de-focused" [Dorfel 1982]. Strongman [1982] conducted similar
studies using seven colors in the display and found the same results,
adding that other colors sampled were acceptable. These additional
colors were sufficient for a pictorial format. For long-term viewing,
the colors chosen to present normal flight information must be
balanced to prevent one color appearing more "harsh and demanding than
the others" [Strongman 1982]. Our study avoided this issue by
utilizing a black path on a white background in sufficient and
consistent lighting.
In many of the studies involving tracking, horizontal movement along a
single, discrete axis was analyzed. The motivation for
one-dimensional movement is based on Fitts' Law in which this is a
requirement. Fitts' Law states that the time, to acquire a target is
logarithmically related to the distance A over the target width W:
t = a + b log2 (2A / W), where a and b are empirical constants
determined through linear regression, and log2 is the log base 2
function [Fitts 1954]. Although the method of input is not
incorporated in Fitts' Law, the "feel" of an input device is extremely
important in determining a device's appropriateness and acceptance in
a particular context of an application. In constrained linear
motions, the mouse, trackball and joystick are the most appropriate
choices for input [Baecker and Buxton 1987] in tracking tasks. In
addition, Buxton [1987] reported that panning is easier with a
trackball than a spring-loaded joystick, because the speed of panning
is proportionally mapped to speed of cursor display; however, if the
tracking must be constrained along a specified path or line, some
unnecessary wrist movements may make the linear motions difficult.
Although MacKenzie and Buxton [1992] unsuccessfully attempted to apply
Fitts' Law to two-dimensional space, the track in our study can be
viewed as a sequence of horizontal and vertical segments, whose
vertices form starting points and targets in a set of single axis
tracking tasks. The target width would simply be the width of the
line. Therefore, an approximation of time to complete the task could
be computed by summing the times to reach one corner from another
corner. It should be noted that the time to trace a horizontal line
may be faster than tracing a vertical line of the same length when
using the joystick method of input. Yamishita and Matsuura [1987]
discovered a superiority of performance in tracking a target in the
x-dimension over one in the y-dimension. They reasoned that the
muscles involved in activating the joystick would have an impact on
the comfortability of movement. Motion in the x-dimension of the
joystick uses the radial and ulnar muscles in the wrist, whereas the
y-dimension involves mainly the dorsi and plantal flexions. The
effects of these motor limitations accompany mental and cognitive
processes in the tracking performance.
Despite the motor limitations, tracking can be improved with practice
to achieve a level of stability [Card, English and Burr]. Sufficient
practice, such that learning is complete (i.e. a stability is
reached), separates the subject's ability component from the subject's
learning component [Alvares and Hulin 1972]. Bliss, Kennedy, Turnage,
and Dunlap [1991] attribute possible correlation between video game
performances and tracking tasks to the practice achieved in using the
input device. The choice of the video game influences the amount of
practice needed - some video games will have no significant effect on
tracking performance. In their study, several video games were
rejected during practice, because performance never reached stability.
In a remotely controlled two-dimensional tracking task with fixed
field of view, the frequency and severity of course deviations errors
is not affected significantly by the type of display system, either
panel-mounted or helmet-mounted.
Continue
Table of Contents