|
Whisper Overview
Simply put, the term "Whisper" refers to an approach to
tracking. More specifically, it refers to a type of acoustic tracker
that we believe will allow accurate body tracking in the presence of
occlusions.
Nick Vallidis, a student whose dissertation focused on Whisper, provides
the following summary.
The full title of my dissertation is
"WHISPER: A Spread Spectrum Approach to Occlusion in Acoustic Tracking".
That should give you some more ideas on what I'm pursuing. Whisper is
just a name that reminded us (any time I use the plural in this document
I am referring to my collaboration with Gary Bishop and/or Greg Welch)
of how this system sounds when it is working. It uses a wide-bandwidth
acoustic signal that is audible to people. WHISPER is not an acronym,
but I write it using small caps (which I can't do here) to signify that
it's the name of the system. The sound is not really like a whisper,
well, unless you were to take out all the sounds that make the words
and leave the hiss. Tune a radio to a channel where there is no station
and you've got a pretty good idea of what it sounds like.
Taking the next set of important words we have spread
spectrum. It is the spread spectrum ideas that are responsible for creating
the wide-bandwidth signal. Except for a commercial system (that might not
even exist) and a master's thesis that only used spread spectrum acoustic
ranging as an application of the work, I don't know of any spread spectrum
acoustic tracking systems. I have seen numerous references to this idea
underwater (once again in the form of commercial systems that I can not
prove the existence of), but not in air. However, this is not the area
where my work is most valuable.
Occlusion is the problem I'm interested in [looking at] with
WHISPER. There has been a large amount of work done to develop head-tracking
devices to allow a computer to generate appropriate images for use in a head-mounted
display. In fact there are two very good systems (UNC's HiBall and Intersense's
Constellation) in existence that accomplish this goal. Now there is a rising
interest in tracking the rest of the body (specifically the hands and legs)
in order to draw accurate avatars in virtual environments in addition to many
other uses. The big issue in this problem space is occlusion. The human body
is capable of such a range of motion that there is no possible placement of
transmitter and sensor pairs such that they would not be occluded at one time
or another. Traditionally, this problem has been dealt with by either ignoring
it, using magnetic tracking devices, or mechanical tracking devices. Optical
and acoustic (all of which have been ultrasonic) systems have ignored this
problem. They simply do not provide tracking results when there are occlusions.
Magnetic tracking devices tend to have low update rates and high latency
(due to the filtering they have to do) as well as accuracy issues in
environments containing metals or ferrous materials. Mechanical devices
have their own troubles including difficulty to don and doff, and tiring
the user or at least altering their motions when using heavy mechanical arms.
The [approach WHISPER takes to] occlusion is the use of
the diffraction of sound. By using sound of lower frequencies (audible frequencies),
WHISPER is able to continue tracking a target when there is an occluder between
the speaker and microphone, although with reduced accuracy.
-Nick (4/23/2002)
The term "Whisper" also refers to the prototype hardware
system built to test this approach. More information about the prototype
hardware can be found on the hardware page.
Two excellent sources of information about Whisper are Nick's defense slides
and his dissertation. See below.
Defense Slides: PowerPoint | pdf Dissertation: pdf
|
