[mirtoolbox] Re: Chroma calculation
- From: "M McVicar" <Matt.McVicar@xxxxxxxxxxxxx>
- To: mirtoolbox@xxxxxxxxxxxxx
- Date: Sat, 3 Apr 2010 20:02:11 +0100 (BST)
Hi Olivier,
this new feature is great, thanks! I was wondering if it is possible to
feed an mirspectrum of this constant-q transform into the mirchromagram
function? Otherwise can I use
get(transform,'Data')
to obtain the (log-scaled) data and compute it myself? Also you said you
made some progress on the optimisation? (I fixed my problem with the time
resolution). Many thanks,
Matt
On Wed, March 31, 2010 9:01 pm, Olivier Lartillot wrote:
> Hi Matt,
>
> Here is an experimental version of mirspectrum.m that allows a new option
> 'ConstantQ' that toggles on the use of Constant Q Transform instead of
> FFT.
> cf. help mirspectrum
>
> This file should be placed inside the @mirspectrum subfolder in MIRtoolbox
> folder.
>
> The problem now is that this new algorithm (based on the initial Brown
> (1991) approach) is quite slow. The use of the Brown and Puckette (1992)
> optimization does not seems to make much sense hence since it would
> require first the computation of the whole FFT, which was what we wanted
> to avoid in the first place, right?
>
> Olivier
>
>
>
> MJ McVicar kirjoitti 30.3.2010 kello 12.29:
>
>> Hi Oliver,
>>
>> indeed this is the advantage to the CQ transform. There is some code
>> written by here:
>>
>> <http://hans.fugal.net/research/cq-octave/cq.m>
>>
>> which is explained here:
>>
>> <http://lad.linuxaudio.org/events/2009_cdm/Saturday/21_Fugal/21.pdf>
>>
>> but thought I'd see if MIRtoolbox is doing this already before
>> implementing myself!
>>
>> Matt
>>
>> --On Tuesday, March 30, 2010 12:18 PM +0200 Olivier Lartillot
>> <olartillot@xxxxxxxxx> wrote:
>>
>>> Dear Matt,
>>>
>>> Currently mirchromagram is based only on FFT, without constant Q
>>> transform, which means the FFT needs to be computed with a quite high
>>> frequency resolution, whereas the use of a constant Q transform would
>>> make the computation more efficient. Did I understand correctly the
>>> advantage of constant Q transform for chromagram computation?
>>>
>>> So it would be nice to add this transform in MIRtoolbox. Does anyone
>>> have
>>> suggestion concerning how to implement it?
>>>
>>> Thanks,
>>>
>>> Olivier
>>>
>>> MJ McVicar kirjoitti 30.3.2010 kello 11.52:
>>>
>>>> Dear list,
>>>>
>>>> is the chromagram computation based entirely on an FFT (after
>>>> filtering)
>>>> or is the constant Q-transform taken implicitly?
>>>>
>>>> Thanks in advance,
>>>>
>>>> Matt
>>>>
>>>> ----------------------
>>>> MJ McVicar
>>>> mm4819@xxxxxxxxxxxxx
>>>>
>>>
>>>
>>
>>
>>
>> ----------------------
>> MJ McVicar
>> mm4819@xxxxxxxxxxxxx
>>
>
>
--
Matt McVicar, Bristol Centre for Complexity Science
mm4819@xxxxxxxxxx
Other related posts:
