[SciPy-User] fast autocorrelation

Pierre Barbier de Reuille pierre@barbierdereuille....
Wed Feb 27 08:44:26 CST 2013


scipy.fftpack has two functions for that: rfft and irfft. They are both
faster than fft/ifft, so maybe you want to use that? Or is it still too
slow?

-- 
Dr. Barbier de Reuille, Pierre
Institute of Plant Sciences
Altenbergrain 21, CH-3013 Bern, Switzerland
http://www.botany.unibe.ch/associated/systemsx/index.php


On 27 February 2013 15:02, Neal Becker <ndbecker2@gmail.com> wrote:

> I'm trying to use fft to do fast auto-correlation.  (Actually, my real
> problem
> is slightly different, but I'll leave that detail out).
>
> The input sequence is complex, so a complex->complex FFT is used.  A
> length 2N
> FFT is used with 0-padding to perform linear correlation.
>
> If now in FFT output we take magnitude-squared, we have a real sequence.  A
> complex IFFT could be used to get the autocorrelation output, but that
> would be
> wasteful.
>
> My question is, is there a trick to more efficiently compute IFFT where the
> input is real?
>
> A couple of thoughts:
> 1. there is a nice trick for real->complex fft (not ifft)
> 2. an fft can be used to compute ifft as  ifft (x) = conj (fft (conj (x)))
>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User@scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.scipy.org/pipermail/scipy-user/attachments/20130227/bd47718e/attachment.html 


More information about the SciPy-User mailing list