i
i
i
i
i
i
i
i
70 4. Building a Practical VR System
However, if the user of the desktop system also needs to experience feed-
back then a haptic output device is required. Most of the latest technological
advances of these devices has been as a direct result of the huge commercial
impact of computer games. What one would class as cheap force-feedback
devices (joysticks, mice, steering consoles, flight simulator yokes) are in high-
volume production and readily available. If building your own VR system,
take advantage of these devices, even to the extent of tailoring your software
development to match such devices. Microsoft’s DirectInput application pro-
gramming interface (API) software library was specifically designed with game
programmers in mind, so it contains appropriate interfaces to the software
drivers of most input devices. We provide a number of input device program-
ming examples in Chapter 17.
There is plenty of scope for the development of custom-built haptic out-
put devices. Flexible pressure sensors, custom motorized articulated linkages
and a host of novel devices might be called on by the imaginative person to
research possible new ways of interaction.
4.2.3 Immersive Interaction
Within the small working volume of the desktop, it is possible to provide a
wide range of input devices and even deliver a believable sensation of force
feedback. On the larger scale of a cave, where freedom of movement and
Figure 4.16. Sensing the position of individual fingers relative to one another is a
complex task. (a) Immersion Corporation’s Cyberglove provides joint-angle data via
a wireless link. (Reproduced by permission of Immersion Corporation, Copyright
c
2007, Immersion Corporation. All rights reserved.) (b) The Fakespace Pinch
Glove has sensors in the fingertips to detect contact between two or more fingers.
(Reproduced by permission of Fakespace Systems, Copyright
c
2007, Fakespace
Systems.)
i
i
i
i
i
i
i
i
4.3. Technology for Motion Tracking and Capture 71
lack of encumbrance with awkward prosthetics is desirable, it is much harder
to acquire input, and virtually impossible to provide comprehensive force
feedback.
Sometimes input may be in terms of spoken instructions, or by pointing
at something or appearing to touch a particular 3D coordinate. To do this, the
VR system needs to be equipped with position detection equipment, which
we discuss in the next section. The cave user may be equipped with a special
pointing device or a glove which can not only have its location detected, but
also when it is combined with a stereoscopic display, it can give a sensation
of touching something (some typical devices in this category are illustrated in
Figure 4.16). These devices do allow the user to be able to gain the illusion of
picking up a virtual object and detecting whether it is soft or hard. But they
still do not let you sense whether it is heavy or light.
4.3 Technology for Motion Tracking
and Capture
The final piece in the jigsaw of immersive VR technology concerns how po-
sition and orientation can be acquired from the real world, relative to a real
world frame of reference, and then matched to the virtual world, which has
its own coordinate system.
With the present state of hardware development, there is still no all-
embracing method of sensing everything we would like to be able to sense
within a real-world space. Some currently developed systems work better on
the small scale, some work better on a large scale, some work more accurately
than others and some sense things others can miss. However, there is usu-
ally a trade off between one desirable feature and another, and so how you
choose to acquire your real-world data will depend on the application you
are working on. One thing is certain: you can make a much better choice of
which technology to use if you know something of how they work, and what
advantages and disadvantages they have. We hope to give you some of this
knowledge within this section.
Following on with the broad idea of real world sensing, we can identity
three themes: motion capture, eye tracking and motion tracking. Although
really they are just the same thing—find the position of a point or points of
interest in a given volume—the present state of hardware development cannot
provide a single solution to address them at the same time. Indeed, we have
i
i
i
i
i
i
i
i
72 4. Building a Practical VR System
broken them down into these themes because a single piece of hardware tends
to work well for only one of them.
Motion capture is perhaps the most ambitious technology, given that it
attempts to follow the movement of multiple points across the surface of
a whole object. Optical or mechanical systems can be used to gather the
position of various points on a body or model as it is moved within a given
space within a given timescale. Later, these template actions can be utilized
in all sorts of different contexts to construct an animated movie of synthetic
characters who have, hopefully, realistic behavior. This is especially important
when trying to animate body movements in a natural way. Since VR systems
are more concerned with real-time interactivity, the emphasis shifts slightly
with the focus directed more towards motion tracking. It is difficult to be
precise about when motion capture differs from motion tracking and vice
versa. However, some examples include:
• Continuous operation. Often motion-capture systems need downtime
for recalibration after short periods of operation.
• Number of points sensed. Motion tracking may need only one or two
points, whereas motion capture may require tracking 20 points.
• Accuracy. A motion-tracking system for a haptic glove may need sub-
millimeter accuracy; a full-body motion-capture system may only need
accuracy to within a couple centimeters.
We shall consider motion capture to be the acquisition of moving points
that are to be used for post-processing of an animated sequence, whilst motion
tracking shall be considered the real-time tracking of moving points for real-
time analysis. So, restricting the rest of our discussion to systems for motion
tracking, we will look at the alternative technologies to obtain measures of
position and orientation relative to a real-world coor dinate system. It will be
important to assess the accuracy, reliability, rapidity of acquisition and delay
in acquisition. Rapidity and delay of acquisition are very important because,
if the virtual elements are mixed with delayed or out-of-position real elements,
the VR system user will be disturbed by it. If it does not actually make users
feel sick, it will certainly increase their level of frustration.
Cu rrently, there are five major methods of motion tracking. However,
their common feature is that they all work either by triangulation or by mea-
suring movement from an initial reference point, or perhaps best of all, by a
combination of the two.
i
i
i
i
i
i
i
i
4.3. Technology for Motion Tracking and Capture 73
4.3.1 Inertial Tracking
Accelerometers are used to determine position and gyroscopes to give orien-
tation. They are typically arranged in triples along orthogonal axes, as shown
in Figure 4.17(a). An accelerometer actually measures force F; acceleration
is determined using a known mass m and Newton’s second law. The position
is updated by p
new
= p
old
+
F
m
dtdt. Mechanically, an accelerometer is
a spring-mass scale with the effect of gravity removed. When an acceler-
ation takes place, it drags the mass away from its rest position so that F = kx,
where x is the displacement and k is a constant based on the characteristic
of the system. In practice, springs are far too big and subject to unwanted
influences, such as shocks, and so electronic devices such as piezoelectric crys-
tals are used to produce an electric charge that is proportional to the applied
force.
A spinning gyroscope (gyro) has the interesting property that if you try to
turn it by applying a force to one end, it will actually try to turn in a different
direction, as illustrated in Figure 4.17(b). A simple spinning gyro does not fall
under the action of gravity; it feels a force that makes it rotate about (precess)
the vertical axis y. If an attempt is made to turn the gyro about the z-axis
Figure 4.17. An inertial tracker uses three gyros and three accelerometers to sense
position and orientation.
i
i
i
i
i
i
i
i
74 4. Building a Practical VR System
(coming out of the page, in the side view) by pushing the gyro down, the
result is an increased Coriolis force tending to make the gyro precess faster.
This extra force
f can be measured in the same way as the linear accelera-
tion force; it is related to the angular velocity = f ( f) of the gyro around
the z-axis. Integrating once gives the orientation about z relative to its
starting direction.
If a gyro is mounted in a force-sensing frame and the frame is rotated
in a direction that is not parallel with the axis of the gyro’s spin, the force
trying to turn the gyro will be proportional to the angular velocity. Inte-
grating the angular velocity will give a change in orientation, and thus if
we mount three gyros along mutually orthogonal axes, we can determine
the angular orientation of anything carrying the gyros. In practice, trad-
itional gyros, as depicted in Figure 4.17(b), are just too big for motion
tracking.
A clever alternative is to replace the spinning disc with a vibrating de-
vice that resembles a musical tuning fork. The vibrating fork is fabricated on
a microminiature scale using microelectromechanical system (MEMS) tech-
nology. It works because the in-out vibrations of the ends of the forks will
be affected by the same gyroscopic Coriolis force evident in a r otating gyro
whenever the fork is rotated around its base, as illustrated in Figure 4.18(a).
If the fork is rotated about its axis, the prongs will experience a force push-
ing them to vibrate perpendicular to the plane of the fork. The amplitu de
of this out-of-plane vibration is proportional to the input angular rate, and
it is sensed by capacitive or inductive or piezoelectric means to measure the
angular rate.
The prongs of the tuning fork are driven by an electrostatic, electromag-
netic or piezoelectric force to oscillate in the plane of the fork. This generates
an additional force on the end of the fork F =
× v, which occurs at right
angles to the direction of vibration and is directly related to the angular ve-
locity
with which the fork is being turned and the vector v, representing
the excited oscillation. By measuring the force and then integrating it, the
orientation can be obtained.
Together, three accelerometers and three gyros give the six measurements
we need from the real world in order to map it to the virtual one. That is,
(x, y, z) and roll, pitch and yaw. A mathematical formulation of the theory of
using gyros to measure orientation is given by Foxlin [8].
The requirement to integrate the signal in order to obtain the position
and orientation measures is the main source of error in inertial tracking. It
tends to cause drift unless the sensors are calibrated periodically. It works very
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.