Re: Experience with the Sliding DFT, anyone?

On Dec 15, 4:11 pm, Andor <andor.bari...@xxxxxxxxx> wrote:
On 15 Dez., 18:55, HardySpicer <gyansor...@xxxxxxxxx> wrote:





On Dec 16, 3:31 am, Richard Dobson <richarddob...@xxxxxxxxxxxxxxxx>
wrote:

hxtasy wrote:
Hello I would like to know if anyone has experience with the sliding DFT
algorithm. It is somewhat similar to the Goertzel algorithm.

All I would like to know is what application this algorithm would be
useful in?

Probably the most unorthodox and extravagant of all possible
applications, but I have been using it for musical (audio) applications,
mainly in its use as part of a full (but very slow!) "sliding phase
vocoder" (SPV):

http://dream.cs.bath.ac.uk/SDFT/index.html

Now fully incorporated in Csound.

Our initial paper on the SDFT was for ICMC2005, which can be found via here:

http://dream.cs.bath.ac.uk/DigitalLibrary/index.php

(use the ICMC link; see also the Dafx08 link for some initial
explorations of a ConstQ form)

We have yet to put our 2007 ICMC paper online, but the first link above
gives access to the slides we used with some sound examples (though it
is far more about the SPV than the SDFT itself).

I am not the one to ask about the maths though - not my area!

Richard Dobson

Do you have a ref for the original sliding DFT paper?

It is just the standard recursive algorithm to compute a running sum
(store the sum in a state variable, subtract the oldest input and add
the newest input) with an additional twiddle factor multiplication.
You'll find tons on the web.

Regards,
Andor- Hide quoted text -

- Show quoted text -

It's a very old technique. You can find it in many textbooks from the
1970's (e.g.: Rabiner and Gold, p. 382-3). Googling "sliding dft'"
found >1600 hits; among them:

http://www.comm.utoronto.ca/~dimitris/ece431/slidingdft.pdf

http://www.ingelec.uns.edu.ar/pds2803/Materiales/Articulos/SlidingDFT_BW.pdf

The second link is the 'dsp tricks and tips' paper from IEEE Signal
Processing Magazine.

It's often used when you don't want all N frequency points that you
would get from a regular DFT or FFT. And, just as with the
conventional DFT, you can compute it for fractional frequencies, and
your data can be any length N.
.

Relevant Pages

  • Re: suggestion: standardize random number generator?
    ... thought was that there are about three useful levels of RNG: ... have any significant statistical anomalies in applications (like ... Twister is a popular algorithm of this class. ... the simpler PRNG's (unless you have special hardware such as the ...
    (comp.lang.forth)
  • Re: Am I the only one with doubts about .NET for commercial apps?
    ... With an algorithm it's relatively easy to ... > I spend relatively little time thinking about coding. ... For some applications that's true. ...
    (microsoft.public.dotnet.general)
  • Re: 3DES Number of security bits
    ... My question is about the strength of 3DES algorithm. ... The parity bits do not become part of the expanded key. ... The thinking is that because of the small block size, the work required to break three-key 3DES is less than its 168-bit key would imply and is much closer to the 112-bits associated with two-key 3DES. ... The main issue at this point is that 128-bit AES is faster and is likely to be more robust than 3DES, so dragging 3DES into new applications seems pointless unless the new applications must interoperate with legacy applications. ...
    (sci.crypt)
  • Re: Experience with the Sliding DFT, anyone?
    ... It is somewhat similar to the Goertzel algorithm. ... applications, but I have been using it for musical applications, ... Do you have a ref for the original sliding DFT paper? ... It is just the standard recursive algorithm to compute a running sum ...
    (comp.dsp)
  • Re: What about turbos?
    ... what percent of the applications (especially from the applications which are built ... focus on "labour saving" languages and language features. ... And the decoupling of algorithm and implementation is not necessarily "a ... Why do we need to be "efficient" in days when memory and CPU power is ...
    (borland.public.delphi.non-technical)