ALISA and biometrics
6 views
Skip to first unread message
Michael Cloppert
Jun 6, 2010, 9:27:33 PM6/6/10
to The ALISA Community
All,
Is anyone familiar with uses of ALISA in biometric systems (e.g.
fingerprint/face/iris ID)? References would be helpful. Broadly, I'm
looking for intersections of ALISA with information security
disciplines; this seems to be the most likely. This is a very active
area of research in my field, yet I have yet to see any mention of
ALISA despite reading quite a bit of papers on the subject, which I
find surprising.
Thanks in advance!
Mike
Is anyone familiar with uses of ALISA in biometric systems (e.g.
fingerprint/face/iris ID)? References would be helpful. Broadly, I'm
looking for intersections of ALISA with information security
disciplines; this seems to be the most likely. This is a very active
area of research in my field, yet I have yet to see any mention of
ALISA despite reading quite a bit of papers on the subject, which I
find surprising.
Thanks in advance!
Mike
Glenn
Jun 6, 2010, 11:38:28 PM6/6/10
to The ALISA Community
Mike,
Good question.
I know Peter, myself, and perhaps others have used the ALISA Texture
Module to do quality assessments on fingerprint images. Peter may have
some examples, but I don't think this work was ever published.
As part of my Shape Module research I did some basic experiments with
fingerprint and finger vein data, but solutions to these applications
have already been commercialized and highly optimized. Iris
recognition is another natural app for something like the Shape
Module, but the Daugman algorithm, which dominates commercial
implementations, is really solid theoretically. It's not rotation
independent like the Shape Module but it's much better for this
particular app where rotation independence isn't necessary..
Face recognition systems have been able to show improved accuracy by
looking at face texture, but here again specialized solutions have
been developed and commercialized.
One of the big challenges in matching some biometric modes (e.g.,
fingerprints, finger veins) is that good clustering solutions have not
been found, so identification matching requires an exhausive search. I
developed a non-metric clustering approach very similar to the Shape
Module's radial feaure token, but in a non-metric feature space
instead of in a 2D image space. The slides on this can be found here:
http://www.biometrics.org/bc2007/presentations/Tues_Sep_11/BSYM/11_Becker_BSYM.pdf
If you're interested, I can send you the paper.
Cheers
Glenn
Good question.
I know Peter, myself, and perhaps others have used the ALISA Texture
Module to do quality assessments on fingerprint images. Peter may have
some examples, but I don't think this work was ever published.
As part of my Shape Module research I did some basic experiments with
fingerprint and finger vein data, but solutions to these applications
have already been commercialized and highly optimized. Iris
recognition is another natural app for something like the Shape
Module, but the Daugman algorithm, which dominates commercial
implementations, is really solid theoretically. It's not rotation
independent like the Shape Module but it's much better for this
particular app where rotation independence isn't necessary..
Face recognition systems have been able to show improved accuracy by
looking at face texture, but here again specialized solutions have
been developed and commercialized.
One of the big challenges in matching some biometric modes (e.g.,
fingerprints, finger veins) is that good clustering solutions have not
been found, so identification matching requires an exhausive search. I
developed a non-metric clustering approach very similar to the Shape
Module's radial feaure token, but in a non-metric feature space
instead of in a 2D image space. The slides on this can be found here:
http://www.biometrics.org/bc2007/presentations/Tues_Sep_11/BSYM/11_Becker_BSYM.pdf
If you're interested, I can send you the paper.
Cheers
Glenn
Michael Cloppert
Jun 7, 2010, 12:20:59 AM6/7/10
to The ALISA Community
Thanks a lot, Glenn. I'll take a look at the slides and get with you
individually if I have any specific questions.
> instead of in a 2D image space. The slides on this can be found here:http://www.biometrics.org/bc2007/presentations/Tues_Sep_11/BSYM/11_Be...
individually if I have any specific questions.
Peter Bock
Jul 4, 2010, 1:42:44 PM7/4/10
to The ALISA Community
Hi Mike:
No, we never did anything substantive with finger prints, although it
is an excellent application for ALISA!!!
Teddy Ko's dissertation presents a powerful method for classifying
exposed areas of human skin. When applied to human faces, the eyes
appear as anomalies (not skin), whihc ought to be useful for security
systems, I would think.
I also did some informal work with lie detection using ALISA, both
remotely and passively. I never applied for funding to continue this
research, but decided to retire instead. :-).
Cheers,
Peter
No, we never did anything substantive with finger prints, although it
is an excellent application for ALISA!!!
Teddy Ko's dissertation presents a powerful method for classifying
exposed areas of human skin. When applied to human faces, the eyes
appear as anomalies (not skin), whihc ought to be useful for security
systems, I would think.
I also did some informal work with lie detection using ALISA, both
remotely and passively. I never applied for funding to continue this
research, but decided to retire instead. :-).
Cheers,
Peter
Peter Bock
Jul 4, 2010, 1:49:23 PM7/4/10
to The ALISA Community
Hi again:
I had a complete opthamological exam a few weeks ago. The doctor
should me the retinal maps his imaging equipment produces. It was
easy, quick, and painless to acquire these images with some automatic
equipment he had. It was very clear to me that the ALISA Geometry
Module or the Component Module would provide a fast and reliable ID
mechanism based on the structures on the surface of the retina.
If you want to discuss any of these ideas further, just let me know!
I had a complete opthamological exam a few weeks ago. The doctor
should me the retinal maps his imaging equipment produces. It was
easy, quick, and painless to acquire these images with some automatic
equipment he had. It was very clear to me that the ALISA Geometry
Module or the Component Module would provide a fast and reliable ID
mechanism based on the structures on the surface of the retina.
If you want to discuss any of these ideas further, just let me know!
Reply all
Reply to author
Forward
0 new messages