|
|
Kathy Lowther and Colin Ware
Faculty of Computer Science, University of New Brunswick
P.O. Box 4400, Fredericton, New Brunswick, Canada E3B 5A3
Email : kathy@omg.unb.ca, cware@unb.ca
Vection is the illusory impression of self motion that can be
obtained when an observer views a large screen display containing
a rotating or translating pattern. To aid in our construction
of an interactive large screen interface to virtual 3D environments,
we conducted studies to determine the factors which induce a sense
of vection. We found that having a foreground frame and a stereo
display increased vection. If subjects moved when the display
was being observed then vection decreased. However, if the perspective
was coupled to their head position while they moved then vection
was restored.
vection, virtual environments
A person viewing a display containing moving objects often feels that they themselves are moving in the opposing direction even though they are not. This illusory self-motion is called vection. It is possible to induce rotational or translational vection. A real world example of this phenomenon occurs when a passenger in a stationary train feels that they have started moving when in fact it is a train on an adjacent track that has begun to move.
A strong sense of vection can be induced by large field-of-view displays with high resolution and an abundance of visual stimuli. This effect is taken advantage of in flight and driving simulators where the illusion of motion through space contributes to the realism of the training environment. Vection may also contribute to a sense of presence in a virtual world, however there is evidence that it may induce motion sickness as a side effect [4].
The ability to convince someone that they are moving in a virtual world may also be an advantage in a 3D interactive graphics system with a large screen display. We are currently building a 3D interaction system that uses a large screen interface like those found in [3,5] but it also has a flying interface that lets us switch between a direct manipulation mode and a flight mode of interaction. It is for this reason we are interested in our ability to induce vection with such an interface.
Perception studies have established certain visual elements which
are important in inducing vection. It has been shown that stationary
objects beyond the moving display hinder the strength of vection,
while a stationary frame in the foreground appears to enhance
the vection experience [2]. It has also been shown that there
is increased vection from objects that appear to be moving at
a distance, as opposed to objects moving in the foreground [1].
Although these visual elements have been shown beneficial in real-world
situations, their effectiveness has not been examined in virtual
environments.
The goal of our first experiment was to establish conditions conducive to inducing vection so that we had an experience of vection assured for the second experiment. The second experiment investigated the effects of stereo and head tracking in vection displays. Due to space restrictions only some of the experimental conditions and results are detailed here.
Experiments were conducted using a large screen rear-projection
display which was approximately 225x180 cm. Computer graphics
were generated by a Silicon Graphics Indigo2 Extreme workstation.
Both rotational and translational vection displays were used. The rotational display was a sphere of longitudinal stripes generated for a viewpoint in the center of the sphere directed towards the equator. This scene was rotated about the vertical axis. The translational display consisted of flow fields of squares situated above and below the subject, analogous to a ceiling and a floor. This was translated either to the left or to the right. The apparent screen velocity was constant and comparable for rotational and translational displays.
In both experiments the subjects were exposed to the vection display for 35 seconds. They were asked to indicate the onset of vection, that is, the first moment that any vection was experienced, by pressing a button. At the end of the 35 seconds, the subject was asked to give a subjective rating for the strength of vection on a scale of 0 to 10, where 0 was no vection and 10 was when users were convinced they were moving.
The first experiment consisted of both types of vection displays shown both with and without a frame. The frame was a set of horizontal and vertical lines situated at screen depth which divided the screen into a 5x5 grid of blocks. This grid is illustrated with the translational display in Figure 1. Each condition was repeated three times for each of the ten subjects in the experiment.
In the second experiment the frame was always present. Both of
the vection displays were shown under four conditions: normal,
stereo, head tracked and stereo with head tracking. To ensure
that head tracking had an effect, the user was required to move.
The subject was instructed to move in a left-right, forward-back
motion one step in each direction. However, this motion had to
be combined with the other three conditions to ensure that any
effect of head tracking was not due to the required motion. This
resulted in a total of eight conditions. Once again, each condition
was repeated three times for each of the ten subjects taking part
in the experiment.
Due to the lack of space in this short report, we report only the onset time data here and not the subjective ratings. The results from Experiment 1 are illustrated in Figure 2. For both rotational and translational displays there are shorter onset times when the foreground frame is present. This agrees with the results of [2]. The results also show a faster onset of vection for the rotational versus the translational displays. Note that in all the above cases shorter onset times implies earlier and stronger vection.
Figures 3 and 4 illustrate the results from the second experiment. We found that a stereo display shortens vection onset times in all cases and that translational displays have faster onsets times than rotational. Results also show that if subjects moved while viewing the display, the onset was slower.
Overall, our results suggest that it is relatively easy to induce the perception of self motion over the course of a few seconds. A major advantage of this is the ability to give observers the sense that they have moved through a virtual space. Future research will involve continued investigation of factors and their effect on inducing vection.
