| Talks, Conference Organization, Projects...
|
| From Shadow Biometrics to Robotic Scaffolds for Tissue
Engineering and Telepresence Mediation...
|
as@adrianstoica.com
+1 626 USA LAB1 |
|
| Shadow
Biometrics |
In 2008 I authored a paper on Shadow
Biometrics. The paper generated quite a bit of press, and some
comments, some uneducated, some plainly unfair....
At that time i created a page to provide some clarrifications - some
comments are found here as well.
The paper dealt with the use of the information in shadow dynamics for
recognition/identification of humans, and classes of human behaviors,
from high-altitude platforms (and possibly/eventually from space).
The original paper "Towards Recognition of Humans and their
Behaviors from Space and Airborne Platforms:Extracting the Information
in the Dynamics of Human Shadows", presented August 6th, 2008 in
Edinburgh, at the 2008 ECSIS Symposium on Bio-Inspired, Learning and
Intelligent Systems for Security (BLISS 2008) is
here. A later paper, with actual results, is here
I only interacted with the New Scientist (NS) magazine journalist
- all others used the NS source, directly, or indirectly
New Scientist article Shadow analysis could spot terrorists by their
walk
A journalist from the New Scientist magazine found the title of my
paper interesting and asked for the paper and some explanations. I
provided these, with the request to see the final magazine article/note
before it was published. I received two drafts on which I provided
comments/corrections - we had a good interaction. However, I did not
see the final version before print - he had his deadlines... The final
version has differences from previous ones. Some of the main
differences may allow for misunderstandings (which happened especially
when the news was taken over by other reporting agencies). Two are
discussed below.
NS: "Nearly seven years after Osama Bin Laden disappeared, US
intelligence agencies are still chasing his shadow. And shadows are
precisely what they should be looking for, says NASA's Jet Propulsion
Laboratory in Pasadena, California."
My comments: This statement is an exageration; I did not say that.
NS: "Shadows provide enough gait data to deduce a positive ID."
My comments: I belive it may (but it also may not) provide enough
data; even for cases when a positive ID is determined, it has
associated with it a given probability of the ID being correct.
NS: "To prove it, he has written software that recognises human
movement in aerial and satellite video footage."
My comments: I introduced the technique and indicated the generic steps
for such a technique/method, but there is no end-to-end software
implementation. I used commercial software to aid me in isolating the
shadows in imagery recorded from a higher floor of a building. Then I
indicated features which could be used for classification/recognition.
Only preliminary work has been performed.
Daily Mail title: Could a spy satellite identify any of us from our
shadow? Daily Mail (DM), took the news and used different sentences,
getting, at times, farther from what was stated.
DM: " The system, being developed by Nasa, is useless once the sun goes
in." .
My comments: NASA is not developing the system. One NASA scientist (me)
performed a study (communicated in the paper). (And by the way, I do
see my shadow at night when I walk on the street - of course, only if I
walk in the light, so such a system, if it existed, would not be
totally useless 'once the sun goes in' :-)
DM: "According to Dr Adrian Stoica of Nasa's Jet Propulsion Laboratory
in California, video from space could provide enough data to confirm a
suspect's identity - as long as details of the person's walking pattern
were on file."
My comments: I never claimed anything about confirming identity
(confirming has a special technical meaning, which is related to
validation). Or, about the walking pattern on file - smart addition
though:-).
See also comments above.
Facts:
Shadows observation enables overhead (remote) biometrics; an
above-the-head observation is transformed into something (a
shape/silhouette) that a system performing gait/gesture recognition can
use - as if it was recorded from ground level. This is the main
novelty, and opens new avenues since there is no alternative technique
today for exploring overhead biometrics.
State-of-the-art gait based identification is in the neighborhood of
60% correct classification rate (CCR) for outdoor recordings (from
ground, direct observation, and other restrictions). I expect that good
quality shadows, observed from above, could provide similar order of
magnitude (of CCR) as if determined from ground observation.
As it is true of all systems, it may be fooled to certain extent (TBD),
by various metods of alteration, e.g. by faking, or by inducing a
modification in/to the shoes, etc. These are issues specifically
related to gait ID, while the focus of my argument is related to
shadows and transforming shadows into silhouettes/shapes that can be
used as input for gait ID/recognition.
Current UAVs could certainly carry the imaging power needed and could
be used to deploy such a system. It is not clear yet how far this is
from being used on the satellites (and no pressing need if a cheaper -
e.g. UAV - solution is feasible now).
The paper was prepared for a conference on techniques for security, and
it appeared normal to focus the introduction of the paper on
applications to security. It can have other applications, such as, for
example finding missing children/persons, etc. One should recognize
that the probability of funding technology development is likely to be
related with the most pressing needs at a given time. Yet, once
developed and available, it can be used in a wide range of
applications. Also, as with any other technology, it can have good
uses, or bad uses; the technology itself is however neutral.
|
|
|
|
