Quacks Like a Duck, Looks Like a Duck, Sounds Like a Duck - Not a Duck
By Erica D. Harrison, CPP
During WWII, Adolph
Hitler’s voice was heard
over the radio, presumably on a
broadcast in one city, while our military
intelligence knew he was actually in
another location. His voice had been
recorded on a ‘wire’ recorder, predating
tape. Thank goodness emerging technology
to record and playback sound
was not solely in the hands of our
enemies…
Today, you might see someone delivering
a speech on video, or presumably
on a live broadcast, and it could be
a very sophisticated and almost indiscernible
fake! Intelligent technology
similar to what we use to analyze
information coming into video surveillance
can also be used to ‘create’ data.
In the case of facial movements, it’s
beyond ventriloquism. Advanced
intelligent software programs divide
the face into the lower half containing
the jaw/lip/mouth/cheek areas and
the upper half with the eyes and
forehead. The two portions are only
re-united after the program ‘learns’
the subject’s speech patterns, inflections,
intonations, facial movements
and quirks.
The ‘subject’ reads a series of words
and phrases that help the program
learn the muscle movements, expressions
and articulations. Once mastered,
the program itself can create words
that the subject never said. The
process of “voice puppetry” uses a
Hidden Markov Model to determine
the audio features of each facial state.
Eventually, the program melds the
face parts back together. Then a relatively
flawless artificial entity looks
like and moves ‘his’ face exactly as the
real person would. You may just see
and hear the unexpected! For more
details, ‘Google’: “Proceedings of
Mirage 2003, INRIA Rocquencourt,
France, March, 10-11 2003.”
We take video of a person moving and
doing maneuvers with his whole body
(walking, sitting, climbing, bending,
reaching, etc.) Computer modeling
programs extrapolate other motions
and ‘create’ a very close approximation
of how that person would perform
other movements, ones the camera
has never captured. A relatively
flawless artificial entity looks like and
moves exactly like the real person. It’s
got a title: Human Motion Signatures
– 3-Mode Factor Analysis.
A quote from M.A. O. Vasilescu, a
leader in the field, comes from papers
summarized in NYU’s on-line media
research lab under Human Motion
Signatures: “We have developed an
algorithm that extracts motion signatures
and uses them in the animation
of graphical characters… For example,
given a corpus of walking, stair
ascending, and stair descending
motion data collected over a group
of subjects, plus a sample walking
motion for a new subject, our algorithm
can synthesize never before seen
ascending and descending motions
in the distinctive style of this new
individual.”
Walks like a duck, talks like a duck,
looks like a duck... Will we always be
able to tell truth from fiction, real from
fake? A plethora of fake forensic evidence
can be at your fingers right now.
The above is condensed and simplified
material on what intelligent [video]
technology delivers right now. We are
only scratching the surface. Meanwhile,
if anyone asks you to read: “Fit”
“Believe” “Zebra” “Dime” and “Jam,”
among a series of phrases while you
are being video taped for any reason,
run like hell.
|
Erica D. Harrison,
CPP, Security
Director at
Guardian Security,
Inc. and President
of AIMS Testing
Inc., has been working
with the New
York City chapter of
ASIS International
since 1985. She has produced and moderated
the seminar programs for the chapter
trade show since 1990, and she writes
extensively on security issues affecting
chapter membership. A former Assistant
Regional VP, Erica holds a bachelor’s
degree from SUNY Stony Brook and a
master’s from Greenwich University. |
>> Return to This Issue's Table of Contents
|
