Mona-
> So I actually do not have a special problem with
something; I
> just wanted to have a productive chat with someone that
> shares my interest in Compressed Sensing!
>
> I am actually working on my thesis on compressive sensing of
> speech. So far everything is working nice practically;
> but I'm having difficulties proving the theory to myself. I know
> it's a new topic, but that actually should make it
> easier to find the proofs and everything; but it's not. You feel
> like the founders of the theory are going all over
> trying to prove the theory, but it doesn't make sense for me:(
> Though the results are amazing and in the same
> direction of the theory!
>
> So if anyone out there is interested in talking about CS
> sometimes it would be my pleasure.
What about this one:
http://people.ee.duke.edu/~willett/SSP//Tutorials/ssp07-cs-tutorial.pdf
...focuses on image processing, but the concepts and theory discussion would
seem to apply to speech 3D spectrogram
representation. There seems to be some speech researchers doing so, for example
replacing sections of noise in the 3D
spectrogram with an estimate of "clean speech" (interpolating), then continuing
with their process (recognition,
compression, etc).
-Jeff