[linux-audio-dev] Prony algorithm?

[Albert Graef]

Christian Schoenebeck wrote:
> I never heard about the "Prony" algorithm, but if it does what you describe 
> then it's easy to implement, isn't it? Have you tried it? Any problems? Or 
> are you concerned about performance?

No, I haven't implemented it yet. That's why I'm looking around for 
available solutions before trying to do it myself. As Fons already 
pointed out, the algorithm isn't trivial at all if you want to do it 
right and in an efficient manner. There are also issues with numerical 
instability when the input is noisy.

In case anyone's interested in this stuff, here's a closer description 
of the problem with some references:

http://www.statsci.org/other/prony.html

I'd really appreciate further pointers about descriptions and efficient 
implementations of that method. I think I read somewhere that netlib has 
a corresponding module, but I wasn't able to locate it.

Alternatively, is there any other spectrum estimation method which gives 
a reasonably good approximation of the main partials of a 
(pseudo-)periodic signal (FFT is too bad in the lower frequencies)? Does 
anyone have any experience (or even code) with using wavelets for that 
purpose?

Cheers,
Albert

-- 
Dr. Albert Gr"af
Dept. of Music-Informatics, University of Mainz, Germany
Email:  Dr.Graef at t-online.de, ag at muwiinfa.geschichte.uni-mainz.de
WWW:    http://www.musikwissenschaft.uni-mainz.de/~ag