[mirtoolbox] Re: MIR chromagram question

  • From: Olivier Lartillot <olartillot@xxxxxxxxx>
  • To: mirtoolbox@xxxxxxxxxxxxx
  • Date: Fri, 27 Aug 2010 10:06:15 -0700

Hi Verena,

Sorry for the late reply to your interesting question.

You should not compute mirchromagram using excessively short frame (such as 30 
ms), because such short temporal window leads to an insufficient spectral 
resolution. The spectral resolution can be artificially improved through 
zero-padding (which is done automatically in MIRtoolbox) but this corresponds 
just to an uninformative interpolation of the deficient spectral information.

In the next version of MIRtoolbox, I will turn back on the display of warning 
message when using such short time windows. (This warning existed before 
version 1.0, but when the automated zero-padding was integrated, we toggled off 
that warning but this was maybe not a good idea.)



Le 16.8.2010 à 7.05, Verena Mattern a écrit :

> Hello list,
> the question I have is more of an observation.
> I used the mirchromagram function once with 'Frame' and the default 100ms
> (0.1) and the other time with 'Frame' and 30ms (0.03). The results I´ve
> got differ alot.
> I´ll try to include the two resulting chromagrams with this this email.
> Otherwise you can find the 30ms chromagram here:
> http://www.pictureupload.de/originals/pictures/160810160043_30ms_chroma.png
> and the 100ms chromagram here:
> http://www.pictureupload.de/originals/pictures/160810160151_100ms_chroma.png
> The 100ms chromagram is correct regarding the input note. The 30ms is not.
> Why does this happen and how can I avoid it and still use 30ms or less?
> Best Regards,
> Verena<30ms_chroma.png><100ms_chroma.png>

Other related posts: