NON UNIFORM FAST FOURIER TRANSFORM
in [Matlab]
|
Prev: fscanf, fprintf
Next: imrect
From: kk KKsingh on 20 May 2010 21:58 Greg Heath <heath(a)alumni.brown.edu> wrote in message <5be3aff6-6123-4262-a71f-79db7a15d2a8(a)a16g2000vbr.googlegroups.com>... > On May 20, 5:02 pm, "kk KKsingh" <akikumar1...(a)gmail.com> wrote: > > TideMan <mul...(a)gmail.com> wrote in message <7535a242-c1b7-4191-9cc6-2380d81d1...(a)t34g2000prd.googlegroups.com>... > > > On May 20, 12:20 pm, "kk KKsingh" <akikumar1...(a)gmail.com> wrote: > > > > JUST Wondering whether any one is working on Non uniform fast Fourier transform > > > > > Oh, you're back.................. > > > Did you take any of the advice you were offered on this subject back > > > in April? > > > Several people suggested several courses of action you. > > > Have you pursued any of these at all? > > > If so, why not show us where you've got to? > > > If not, why not? > > > > I was able to get the same result as DFT from my code ! But now i need to go back inverse i am going some thing wrong ....MY NFFT and DFT is same so no issues with it ...But inverse is not easy for me > > I have responded to so many of your posts I have lost track. > The answer to this question is in one of my previous replies. > > The bottom line is: > > If you assume the nonuniformly sampled time function > can be recovered from the traditional IDFT numerical > integration formula, the spectrum should be obtained > from the time function using least squares, not the DFT > numerical integration formula. > > However, if the time and frequency sampling are both > uniform and use the same number of points, the DFT > is the least-square solution. > > Hope this helps. > > Greg Hey Greg! Well, Its due to you i was able to write the code ! In Signal processing for non uniform signals people have adopted strategy of using DFT , and do the reconstruction using least sqaure in frequency domain ! Some used the high amplitude value from the DFT spectrum and reconstruct the signal ! My NFFT algorithm only replaces this DFT nothing else making the steps faster ....but what i was not able to do is constructing the inverse solver, what i mean is step ahead of DFT .....there are plenty of paper available in literature regarding this : http://cat.inist.fr/?aModele=afficheN&cpsidt=1803865 Also http://www-user.tu-chemnitz.de/~potts/nfft/ In the link above Dr potts and Dr stefan build the fantastic libray ! for this purpose only ! Although my code is slower than their but its working for now..except the inversion part ! Have a look on the Inversion part of these people They are getting the approximate spectrum which is also known as DFT ........AND NFFT is we use their fast algorithm...... And later they reconstruct it using Least square I tried to use their libraries but i found it difficult to use, also for understanding i written my own code i have spectrum now from there i am doing some thing wroong Thanks kk |