[SciPy-User] How to calculate Yulewalk with scipy.optimize.leastsq

[josef.pktd at gmail.com]

On Thu, Jan 19, 2012 at 8:25 AM, klo uo <klonuo at gmail.com> wrote:
> Thanks Fabrice, I'll check talkbox later today and reply if that approach is
> working for me
>
> Josef, thanks for your snippet, thou I have no idea how to use it

a quick guess is that it's the same as the denf function, but given
only the one sided autocovariance function.

> I tried to translate Matlab's yulewalk.m, and it kinda works but with an
> issue. Code is here: http://paste.pocoo.org/show/537224/
>
> Main problem is on lines 79-80 when impulse response is computed. FFT/IFFT
> functions result is different. Everything to that point gives correct result
> in respect to Matlab function. I don't know why is that, is it precision in
> question or else, but I had enough hard time to produce linked and not so
> representative code.

No help from my side, I don't understand the code.

I'm curious why they use log and exp during fft
ifft(exp(fft(hf)))
The only other mention of this that I found is on the octave mailing list.

numf looks interesting if I find a reference to avoid license problems.

Josef

>
> I attached example image to show how results differ slightly.
> Any comment is appreciated
>
> Cheers
>
> _______________________________________________
> SciPy-User mailing list
> SciPy-User at scipy.org
> http://mail.scipy.org/mailman/listinfo/scipy-user
>