Underlying Technologies Part One: Hardware Mark Green School of Creative Media
Download
Report
Transcript Underlying Technologies Part One: Hardware Mark Green School of Creative Media
Underlying Technologies
Part One: Hardware
Mark Green
School of Creative Media
Introduction
Take
a look at the hardware technologies
that make it all work
an overview, survey or what’s available,
may come back to details later
two types of hardware:
output
input
Output Technologies
Numerous
modes:
visual
sound
tactile/force
smell
examine
both commercial and research
technologies
Visual
We
tend to concentrate on visual, this is
our dominant sense
aim: fill our visual field of view with
computer generated graphics
standard monitor doesn’t work, only
covers a portion of the visual field, still a
useful tool
Head Mounted Display
Started
with Head Mounted Display (HMD)
two display screens, one for each eye,
mounted on head
images change as user moves
display responds to user’s motions, move
head to view complete 3D environment
not perfect: heavy, low resolution, etc
Projection Based Displays
A number
of display technologies based
on projection
active area of research:
rapidly evolving projection technology
rapid decrease in price
easier
to use and higher quality than HMD,
but requires more space
Projectors
Three
main projection technologies:
CRT
LCD
DLP
CRT
is the oldest technology, based on
three CRT tubes: red, green and blue
low brightness, around 240 lumens
CRT Projectors
Main
benefits:
high resolution
very fast
very flexible
Problems:
hard to maintain, heavy
dead technology, no evolution
expensive
LCD Projectors
LCD
panel with bright projection lamp
LCD panel: grid of cells containing liquid
crystal
when voltage applied to cell crystals
change optical properties, produces image
main problem: crystals don’t change fast
enough
brightness: 600 - 2400 lumens
LCD Projectors
Benefits:
cheap
small, light, easy to maintain
reasonably bright
Problems:
not very fast, 20 - 30 Hz
fixed resolution
DLP Projectors
Digital
light processor: micro-mirrors built
on a chip, voltage applied to mirror causes
it to tilt
shine light on DLP, mirror direction
controls how much light is reflected
a developing technology
brightness: 2,000 - 10,000 lumens
DLP Projectors
Benefits:
very bright
reasonably fast and easy to maintain
developing technology, will get better and
cheaper
Problems:
currently quite expensive
fixed resolution
Front Vs. Back Projection
Projector
can be in front or behind the
screen
back projection: viewers don’t shadow the
screen, but requires a lot of space
front projection: saves a lot of space, but
viewers can shadow the screen
a space Vs quality trade off
Throw Distance
Distance
from projector to screen for best
quality image
varies with projector and optics, rough
estimate: twice the largest dimension of
the image
can be reduced by using mirrors, fold the
optical path, need high quality mirrors and
accurate placement
CAVE
CAVE:
room with projected stereo
graphics on walls, user position tracked
Why?
Very high resolution
less to wear
multiple users
access to real world objects
Walls
Project
onto a single wall, either flat or
curved
may use multiple projectors to give higher
resolution, need edge blending
can be quite large, over 25’ long
usually stereo, but not head tracked
used extensively in oil industry
Desks
Back
project onto a table top, usually
stereo with head tracking
good for fine manipulations
restricted to a few users, can’t do good
stereo with more than 2 users
single projector, so cheaper than a wall or
cave
Sound
Before
we can discuss sound hardware
we need to understand how we hear
there are a number of cues that we use to
locate sound sources
one is the volume, sound volume
decreases with the square of distance
for many sounds we know how loud they
normally are, can estimate distance
Sound
Another
sound cue is based on comparing
results for our two ears
unless the sound is directly in front or
behind us, there will be a time difference
between the ears
there will also be a slight difference in
volume
the main idea behind stereo sound
Sound
The
most important cue comes from the
shape of our outer ear
the outer ear filters sound, filtering
depends upon sound direction
we learn the filtering performed by our ear,
can then judge where sounds are located
we only need one ear for this, two ears
really aren’t that important
Sound
Outer
ear filtering called HRTF - headrelated transfer function
HRTF can be measured by placing a small
microphone in the ear, measure sound
intensities for sound coming from different
directions
use HRTF to filter sound, feed to each ear
separately using headphones
Sound
If
the listener and sound source are
moving relative to each other get Doppler
effect
sounds moving towards us increase in
frequency, sounds moving away decrease
in frequency
this is noticeable at around 30 km/h, so its
something we could expect in our virtual
environment
Sound
How
do we implement this?
Drop HRTF, the rest is fairly easy for
simple acoustic environments
given listener position in VE, and positions
of sound sources, only need to do some
simple filtering, either CPU or sound card
output then goes to headphones
Sound
If
using speakers its more complex
need to know the position of listener in
physical space, plus position of speakers
generate separate audio feed for each
speaker
typically needs at least 4 speakers, but
depends upon size of real environment
Sound
For
HRTF we need to use headphones,
also need to have HRTF for listener,
generic HRTF can be used
can be done on modern computers without
much problem
assume simple acoustic environment, no
walls, no sound reflection, etc
Sound
Simulating
acoustical environments is
more difficult
typically need special purpose hardware
for real-time, usually multiple DSPs
restricted to a small number of sound
sources and a small number of sound
reflectors
Tactile and Force
Related
sensations, both involve hands
and touch
force involves objects acting on the user’s
hands, basically a push against the hand
the user must be holding the device that
provides force output, typically both an
input and output device
force feedback joystick
Force
Force
feedback joystick, like a normal
joystick, except motors attached to the
stick
stick can be moved under computer
control
react to user’s motion, if the hand strikes a
wall stick stops moving
stick shakes when gun fires, etc
Tactile
Sense
of touch, mainly research
try to stimulate tactile sense organs in
fingers and hand
usually some form of vibration:
film in front of loud speaker
array of small pins, each pin can vibrate or
move slightly
Smell
Both
easy and hard, not a standard
technology!
Fairly easy to generate smells, just need
the right chemical in small vials
to generate smell, open the vial and use a
small fan to push smell towards the user
opening and closing vial under computer
control
Smell
Real
problem is stopping the smell
need some way of getting rid of smell
when no longer needed
could use a large fan, probably interferes
with the experience, may make too much
noise
also need some way of mixing smells
predictably
Input
To
interact with a virtual environment there
needs to be some way of sensing our
actions
VEs are 3D, so we need to be able to
enter 3D data:
position
orientation
we
also want to manipulate objects, so
there are other things we might want
3D Data
There
are a number of devices for 3D data
an important sub-class are trackers, used
to determine position and orientation of
body parts
trackers are used by most output devices,
so we will start with them
first look at some important properties,
then look at hardware
Trackers
Latency:
most important property
time from when user moves until position
reported to computer
system latency: time from when user
moves until motion reflected in display
must be kept less than 0.1 second,
otherwise users become very sick
Trackers
Noise:
when the tracker is held still is there
a variation in its output
large amounts of noise decrease usability
Accuracy: does the tracker accurately
report its position
warped spaces are hard to work in, handeye coordination is hard if devices aren’t
accurate
Electromagnetic Trackers
Source
emits an electromagnetic field,
sensor detects field
computes position and orientation from
field measurements
reasonably accurate, but expensive
limited range, problems with metal objects
sometimes add buttons for interaction
Hybrid Trackers
Based
on combining two existing tracking
technologies
Inertia trackers are very fast, based on
measuring accelerations
Perform very well over short time periods,
several seconds, but drift over longer
periods
Hybrid Trackers
Ultrasound
is slow and requires many
sound sources for accuracy
Combine the two technologies:
Use inertia tracking over short time periods
Use ultrasound tracking to correct the inertia
tracker periodically
Produces
system
a fast and accurate tracking
Hybrid Trackers
Advantages:
Fast
Accurate
Low noise
Problems:
Expensive
Immature technology, still needs some work
Video Tracking
Based
on using one or more video
cameras to determine user’s position
Full 3D requires multiple cameras, can do
some things with a single camera
General approach involves multiple
cameras in different locations, capture an
image from all of the cameras
Video Tracking
In
each image find the user and then the
points on the user being tracked
Identifying tracked positions is hard, may
not even be visible
Special optical markers often used to help
identify tracked positions
Extract position information from image,
this is usually 2D data
Video Tracking
Combine
the information from all of the
cameras to get 3D position
Hard to do in real time, especially with
high accuracy
Quite often simplify the problem by using
special backgrounds, special lighting, etc
Video Tracking
Single
camera systems can be used in
special situations
Face tracking: know what we are looking
at and over a limited range
With standard webcam can do a good job
of determining positions of facial features
Optical Tracker
Problem:
suit-up for interaction, must wear
devices, cables on floor, etc
casual use is difficult, just walk into Cave
and use it, nothing to wear
observations:
really a 2D problem, users don’t change
height
high accuracy not necessary for casual
interaction
Optical Tracker
Floor
level laser beam
camera with filter, tuned to laser frequency
camera mainly sees laser light, doesn’t
see graphics
user’s legs interrupt laser beam:
position of break gives one axis
height of break give the other
under
$400 to build
Optical Tracker
3D Data
A number
of devices can be used for
position and orientation, but aren’t trackers
A good example of this is a 3D joystick
Twisting the control stick is used as the
third dimension
Reasonably good, but all 3 dimensions
aren’t treated equally
Pucks
Similar
to joysticks, but easier to use in 3D
round flat puck instead of control stick, can
be moved in x, y and z as well as rotated
about the same price as a joystick, hasn’t
really caught on as a game device, so not
as easy to find
Mechanical Trackers
Looks
like a robot arm
user controls the end of the arm, can
move it freely in 3D space
end of arm looks and feels like a pencil
move the end like a pencil, can draw in 3D
space
can trace out small 3D objects
Mechanical Trackers
Body Gestures
Trackers
give us a few points on the body,
but don’t give complete gesture
information
To grasp a virtual object would like to
move hand to the object and grab it
Need to know what the hand is doing, are
the fingers grasping the object
Several approaches, none are satisfactory
Gloves
The
“Hollywood” VR device
not extensively used now, more of a niche
device
hard to accurately recognize gestures
gloves tend to get dirty and can’t be
washed
takes time to put on
very expensive
Body Suits
Similar
technology to gloves, but cover the
entire body, look like diving suit
Several attempts, but never quite worked
Difficult to get on
Hard to calibrate, determine what the body
is doing
Large amounts of data that is hard to deal
with
Video
Video
techniques have been used for
hand gestures
Work well if hand doesn’t move, much
more difficult if user can move around
Stationary user: one or two cameras can
determine most hand gestures in real time
Speech
Hands
free interaction, extra channel of
communications
Good quality low cost speech recognition,
accuracy > 95%
Some attempts to use in VR, with limited
success
Speech recognition works best in a quiet
environment
Speech
Unfortunately,
many VR systems are not
quiet enough
Main problem is fan noise, doesn’t bother
us much, but causes problems for speech
recognition
Would like to use wireless microphone,
also causes problems for speech
recognition