Académique Documents
Professionnel Documents
Culture Documents
7: Input
p Devices and
Tracking Technologies
Interaction and VR
“ Interaction with a virtual world is a key
ingredient of a VR experience.
IIndeed, if the display of a virtual world
d d if th di l f it l ld
does not respond at least to a user’s
physical movement, then it is not
considered virtual reality.”
– ‐ Sherman and Craig, 2003
Interaction Methods
• Should be easy and intuitive
• Are often unnatural (but humans adapt)
• May or may not mimic real life
• Often rely on metaphors and people’ss
Often rely on metaphors and people
familiarity with them
1
28/11/2010
Interaction Tasks
1. Manipulation
• Comprising selection and
action (sometimes
performed simultaneously)
2. Navigation
• Travel
– How the user moves
through space
• Wayfinding
– How the user knows
where they are and where
they are going
Interaction Tasks
3. Interacting with others
• Shared environments
• Collaborative environments
4. Interacting with the VR
system
• e.g. Changing world
parameters
Tracking Technology – Performance
Criteria
The following come from S. Kalawsky – The Science of Virtual Reality:
• Static accuracy: The ability of the tracker to determine the coordinates of
a position in space. This is the value without averaging or similar filters
applied
• Dynamic Accuracy: The accuracy of the system as the sensor is moved.
Highly dependant on the integration period of samples, if a long
integration period is used then the dynamic accuracy may be very high
integration period is used then the dynamic accuracy may be very high
• Latency: The latency is the time taken to get new data from the sensors
including how quickly the sensors respond to changes in position.
• Update rate: The time taken to process the data from the sensors and
calculate a position. If running faster than the capture (latency part) it may
reuse old data (static accuracy may be high). A high update rate alone isn’t
necessarily an advantage.
• Phase Lag: The total time delay of the tracker system, i.e. latency + update
rate or the true age of an atom of data as it leaves the tracker system.
• Registration: The correspondence between actual position and reported
position. Basically calibration both initial and over time, drift.
• Signal to noise ratio: What it says, signal relative to the background noise.
2
28/11/2010
Tracking Technology – Performance
Criteria
But consider also:
• Degrees of freedom per target
– Normally a 3 DoF tracker will measure EITHER position
x,y,z OR orientation roll, pitch, yaw
– A 6 DoF tracker will measure both position and
orientation
• Number of simultaneous targets
– Some tracking system can’t track many targets at once
– Others must reduce performance for every additional
target due to limited bandwidth or similar
• Workspace
– Some trackers have a fixed maximum workspace,
others may lose performance as workspace increases
Mechanical Trackers
• Includes joystick and mouse (older),
• More usually cables and kinematic links
with sensorised joints
• Pros:
– Good
Good accuracy, responsiveness,
accuracy responsiveness
registration and robustness.
– Very low latency
• Cons:
– Tend to have a limited range of operation.
Physical links have a fixed length, cables
get tangled if rotated
– Low transparency.
Optical Trackers
• Active: LEDs
• Passive: fiducials, retroreflective markers, natural markers
in the environment
• Inherent trade of between accuracy and range of operation
• Can get either image processing or fixed transducer types
• Can be inside
Can be inside‐out
out or outside
or outside‐in
in i.e. receivers are fixed or
i.e. receivers are fixed or
mobile
– Outside‐in has the advantage that a
number of cameras can be used to
track a single target which increases
accuracy and improves line‐of‐sight
– Inside‐out has the advantage of being
maximally sensitive to orientation
• Image processing systems require
controlled light conditions
3
28/11/2010
Optical Trackers
• Pattern recognition systems used to be less
attractive because the computation was too high
but processing time is minimal now
• Pros:
– High data rates (much faster than sound), great for
real time
real time
– Can be made simply, easy access to technology
– Passive systems very scalable (many targets)
• Cons:
– Line of sight
– Ambient noise (even for IR)
– Non‐uniform performance across axes, often depth
less accurate (same reason as humans)
Magnetic Trackers
• For a while magnetic trackers were the most widely used solution though
not any longer
• AC type (7‐14Khz usually)
– Uses three‐axis magnetic dipole source and three‐axis magnetic sensors
– Metallic or other ferrous materials in the environment can induce distortions
as AC causes eddy currents (in newer technology this effect is greatly reduced)
• DC type (pulsed)
– Three axis magnetometers or hall effect sensors (mutually orthogonal)
Three axis magnetometers or hall effect sensors (mutually orthogonal)
– Leave a short time delay for eddy‐currents to die out
– Affected by any permanent magnetic field (earth, power outlets, motors etc)
• Can also use Earth magnetic field intentionally for yaw measurement
• Pros:
– Very responsive, low latency
– Can track multiple targets though update rate drops (probably not much more
than 10 at once for real time)
• Cons:
– Lose accuracy and registration rapidly as working volume increases (S/N ratio
drops)
Acoustic Trackers
• Use ultrasound either time‐of‐flight or phase coherent measurement
techniques.
• Usually target has emitters and receivers are in fixed locations. Can have lots of
receivers to maximise line of sight
• Phase systems:
– Better performance because of higher data rates
– Can operate continuously rather than having to wait for an echo to return
– Don
Don’tt measure absolute position, they measure change and are susceptible to drift
measure absolute position they measure change and are susceptible to drift
• Time‐of‐flight:
– More vulnerable to acoustic noise.
• Pros:
– Good accuracy, responsiveness, robustness and registration (if workspace kept
small)
– Can have multiple targets (but at a reduction in update rate)
• Cons:
– Occluded sensors have a big impact on performance
– Acoustic properties affected by environment (humidity, temperature, pressure) but
can be compensated for
– Position accuracy better than orientation (unless target has big gap between
emitters)
4
28/11/2010
Inertial Trackers
• Combine multi‐axis gyroscopes (angular velocity) and accelerometers
(MEMS)
• Measure roll and pitch accurately using accelerometers (earth gravity
field)
• Yaw must be integrated from gyro measurement
• Position x,y,z much less accurate. Accelerometers are double integrated to
get position
• Errors accumulate with time: e.g. error of 0.0098m/s2 (1 milli‐g). error is
Errors accumulate with time: e g error of 0 0098m/s2 (1 milli g) error is
double‐integrated: 0.0098 t2/2 = 4.5 meters after 30 seconds
• Filtering techniques can account for large part of white noise (evenly
distributed noise)
• Can’t account of Gyro Bias Drift – fundamental limitation of gyros
• Pros:
– Unlimited working volume
– No line‐of‐sight issues
– Tilt sensing very accurate
– Highly responsive
• Cons:
– Yaw and position has measurement drift. Errors accumulate quickly over time.
Sensor Fusion in Trackers
• Rarely actually use one technology in isolation
• More common to get information from multiple
sources and fuse the data to get a better estimate
• Common sensor fusion techniques include Kalman
filtering and Bayesian networks
• Inertial sensors for orientation are usually combined
with a 3 axis magnetometer to adjust for yaw drift
• Inertial sensors for position are very often combined
with optical or acoustic sensors to account for drift
• Other sensing technologies are used such as force
sensors, bend sensors, capacitance and GPS but
normally in combination with the main ones those
already discussed
Input Devices ‐ Examples
• Standard keyboard and mouse
• Joysticks
• Wired gloves
• Motion Capture Suits
• Force balls
• 3D/6 DOF wands and mice
• Speech recognition
• Biosensors
• Eye/gaze tracking
5
28/11/2010
Input Devices – Wired Gloves
PinchGlove
DataGlove
CyberGlove CyberGrasp
Input Devices – Force Balls
SpacePilot
SpaceBall
SpaceMouse
SpaceTraveler
Input Devices – Wands and Mice
Wanda 3D Pointer
Intersense Wand (used in the Reading CAVE)
6
28/11/2010
Input Devices – MS Kinect
• Wide variety of input types:
capable of gesture, face and
speech recognition
• Normal camera plus 2 cameras for depth sensing and 4
microphones
• Depth sensing works by projecting a pattern on the
Depth sensing works by projecting a pattern on the
scene (presumably in IR). Two cameras for depth are
sensitive to pattern. Pattern is denser the further away
it is
• Normal camera + depth sensing used together to
isolate individual users and find facial characteristics
and gestures
• Speech recognition can isolate separate voice
commands (probably using something like ICA) and
select which person is talking from camera feed