MATLABS algorithm for underdetermined (blind) source separation. Where to download?

Hello

I have two unknown speech sources s1 and s2 and one observation y:

y[k]=c1*s1[k]+c2*s2[k];

where c1 and c2 are constants.

Where can I download a matlab algorithm that performs source separation
for under-determined (under-complete) systems ? I need some matlab algorithm
which is able to estimate s1 and s2 based on y.

Thanks in advance...

Did you find anything?  I've been looking for the same thing.

Richard

Lars Hansen wrote:
> Hello
> 
> I have two unknown speech sources s1 and s2 and one observation y:
> 
> y[k]=c1*s1[k]+c2*s2[k];
> 
> where c1 and c2 are constants.
> 
> Where can I download a matlab algorithm that performs source separation
> for under-determined (under-complete) systems ? I need some matlab algorithm
> which is able to estimate s1 and s2 based on y.
> 
> Thanks in advance...
> 
> 
> 

> I have two unknown speech sources s1 and s2 and one observation y:
>
> y[k]=c1*s1[k]+c2*s2[k];
>
> where c1 and c2 are constants.
>
> Where can I download a matlab algorithm that performs source separation
> for under-determined (under-complete) systems ?
> I need some matlab
> algorithm which is able to estimate s1 and s2 based on y.

Homework for a community college class  ?
Must be pretty advanced community college :-)

Nope...not so far...it's probably impossible...Maybe some statistical 
approach is the answer....in my case the two source signals s1 and s2 are 
correlated:

y[k]=a*s1[k]+b*s2[k] where a,b are unknown constants and s2[k]=unknown 
function of s1[k]

So maybe the correlation can be used for something?


> Did you find anything?  I've been looking for the same thing.

Did you try looking on the internet? Sheesh...

http://www.tsi.enst.fr/~cardoso/guidesepsou.html

thanks for the link, but I don't see any code/paper dealing with the problem 
I am trying to solve?

-------------
"spasmous" <spasmous@yahoo.com> skrev i en meddelelse 
news:1131736652.625835.38960@g43g2000cwa.googlegroups.com...
> Did you try looking on the internet? Sheesh...
>
> http://www.tsi.enst.fr/~cardoso/guidesepsou.html
> 

Lars Hansen wrote:
> thanks for the link, but I don't see any code/paper dealing with the problem
> I am trying to solve?
> 

OK, my bad. The jade code only works for sensors >= sources.