DSPRelated.com
Forums

Independent Component Analysis (ICA)

Started by adityar7 November 19, 2006
Hi,

I have been reding about ICA as a method of extracting a new bases for
data. I am wondering if anyone has tried performing ICA on a set of
images. I am particularly interested in its applications to face
detection/recognition. Right now I'm just doing some research and
haven't really decided on any concrete ideas, so I just have some
simple questions.

Has anyone here performed ICA on a set of images? Do you get something
like the "eigenfaces" that you get by performing PCA? Also, is there an
existing C library that will allow me to do the above (I know that
there's a MATLAB library but unfortunately I don not have access to
MATLAB, only Mathematica, whcih doesn't seem to have an ICA library).

Right now, I just want to see what I get by applying ICA to a set of
faces. Any help would be greatly appreciated. Thanks!

adityar7 wrote:

> Hi, > > I have been reding about ICA as a method of extracting a new bases for > data. I am wondering if anyone has tried performing ICA on a set of > images. I am particularly interested in its applications to face > detection/recognition. Right now I'm just doing some research and > haven't really decided on any concrete ideas, so I just have some > simple questions. > > Has anyone here performed ICA on a set of images?
Yes.
> Do you get something > like the "eigenfaces" that you get by performing PCA?
Yes, pretty much exactly like PCA. In fact, PCA is a preprocess stage for ICA.
> Also, is there an > existing C library that will allow me to do the above (I know that > there's a MATLAB library but unfortunately I don not have access to > MATLAB, only Mathematica, whcih doesn't seem to have an ICA library). >
FastICA is available atleast for Matlab, R and C++ http://www.cis.hut.fi/projects/ica/fastica/
> Right now, I just want to see what I get by applying ICA to a set of > faces. Any help would be greatly appreciated. Thanks!
-- Jani Huhtanen
adityar7 wrote:

> Has anyone here performed ICA on a set of images? Do you get something > like the "eigenfaces" that you get by performing PCA?
I hope I never see one of those in the mirror ... :-).
Andor wrote:
> adityar7 wrote: > >> Has anyone here performed ICA on a set of images? Do you get something >> like the "eigenfaces" that you get by performing PCA? > > I hope I never see one of those in the mirror ... :-).
Whose face would you rather see? :-) Jerry -- Engineering is the art of making what you want from things you can get. �����������������������������������������������������������������������

On 20 Nov., 18:06, Jerry Avins <j...@ieee.org> wrote:
> Andor wrote: > > adityar7 wrote: > > >> Has anyone here performed ICA on a set of images? Do you get something > >> like the "eigenfaces" that you get by performing PCA? > > > I hope I never see one of those in the mirror ... :-).
>Whose face would you rather see? :-)
I can hardly imagine that an eigenbasis decomposition of my face is something that I can handle looking at, especially in the morning :-).
Andor wrote:
> > On 20 Nov., 18:06, Jerry Avins <j...@ieee.org> wrote: >> Andor wrote: >>> adityar7 wrote: >>>> Has anyone here performed ICA on a set of images? Do you get something >>>> like the "eigenfaces" that you get by performing PCA? >>> I hope I never see one of those in the mirror ... :-). > >> Whose face would you rather see? :-) > > I can hardly imagine that an eigenbasis decomposition of my face is > something that I can handle looking at, especially in the morning :-).
The word was "eigenface". "Eigen" means "self". Isn't an eigenface what one expects in a mirror? "Eigengesicht" would be linguistically truer, but we mix languages all over, so no matter. Jerry -- Engineering is the art of making what you want from things you can get. &#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;
Jerry Avins schrieb:

> Andor wrote: > > > > On 20 Nov., 18:06, Jerry Avins <j...@ieee.org> wrote: > >> Andor wrote: > >>> adityar7 wrote: > >>>> Has anyone here performed ICA on a set of images? Do you get something > >>>> like the "eigenfaces" that you get by performing PCA? > >>> I hope I never see one of those in the mirror ... :-). > > > >> Whose face would you rather see? :-) > > > > I can hardly imagine that an eigenbasis decomposition of my face is > > something that I can handle looking at, especially in the morning :-). > > The word was "eigenface". "Eigen" means "self". Isn't an eigenface what > one expects in a mirror? "Eigengesicht" would be linguistically truer, > but we mix languages all over, so no matter.
I guess seeing an eigenface is better than seeing a fremdface in the mirror. But still ... it just sounds scaray! :-)
Jerry Avins <jya@ieee.org> writes:

> "Eigen" means "self".
I always thought eigen meant "proper." Not true, or is this an alternate meaning? -- % Randy Yates % "Maybe one day I'll feel her cold embrace, %% Fuquay-Varina, NC % and kiss her interface, %%% 919-577-9882 % til then, I'll leave her alone." %%%% <yates@ieee.org> % 'Yours Truly, 2095', *Time*, ELO http://home.earthlink.net/~yatescr
Randy Yates wrote:
> Jerry Avins <jya@ieee.org> writes: > >> "Eigen" means "self". > > I always thought eigen meant "proper." Not true, or is this an > alternate meaning?
It's true if you construe "proper" an in "property". As a simple translation, that's archaic. "My own sister" translates to "meine eigene Schwester." (Ask Google if you doubt it.) :-) Jerry -- Engineering is the art of making what you want from things you can get. &#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;
Jerry Avins skrev:
> Randy Yates wrote: > > Jerry Avins <jya@ieee.org> writes: > > > >> "Eigen" means "self". > > > > I always thought eigen meant "proper." Not true, or is this an > > alternate meaning? > > It's true if you construe "proper" an in "property". As a simple > translation, that's archaic. "My own sister" translates to "meine eigene > Schwester." (Ask Google if you doubt it.) :-)
The German "eigen" seems to map to the Norwegian "egen", which can have meanings like "own" or "self" as above. Other translations include "stubborn", ("han er sv=E6rt egen", "he is very stubborn"), "particular" or "certain" as in "hun har en egen sjarm", "she has a certain / particular charm". My understanding of the the term "eigenvalue" (N. "egenverdi") has always been "self value". Rune