[Clam-devel] CombFilter and GenderChange

[David García Garzón]

sure, my change was
$ svn diff -r 9924 CLAM/src/Processing/Transformations/SMS/SMSPitchShift.cxx    
Relevant line is 93

If you remove the 'false', you can easily reproduce the bug on the 
NetworkEditor. By synthesizing the output you will get an U shaped output 
residual audio that is listened as a high frequency noise on the output. If 
you take the residual directly from the analysis, (that's what i am currently 
doing with the false)  then you don't get the noise but some pre-shift 
harmonics are still present on the output.

Maybe CombFilter is not taking into account some spectrum format flags? Or we 
are setting them badly on the incomming data?

Another issue i don't understand is why comb filter is filtering target 
frequencies (i*fundamental*amount) instead source frequencies (just 
i*fundamental).


On Dissabte 14 Abril 2007, Xavier Amatriain wrote:
> David, can you give me more details on what files you refactored and
> where I should start looking to help with
> this? My sandbox is a little out of sync so any pointer you can give
> will help me in debugging this. Feel free to
> contact me off-list.
>
> David García Garzón wrote:
> > Yes sorry. The Combfilter i disabled was the one on the PitchShift, and
> > yes, it was suffering this bug. I added an 'if (false ...' on line 93 to
> > disable it. The result is better than the result of enabling it but i
> > listen a background alias sound that i think it was intended to be
> > removed by the comb filter, wasn't it?
> >
> > On Divendres 13 Abril 2007, Xavier Amatriain wrote:
> >> I assume you are talking about the CombFilter in the PitchShift... is
> >> the PitchShift affected by that bug? Is it still working
> >> correctly?
> >>
> >> David García Garzón wrote:
> >>> Xavi, on the list, but this mail is addressed to you.
> >>>
> >>> When i did the GenderChange fixes for realtime i had to disable the
> >>> CombFilter it contains. It generated a lot of sharp noise on the
> >>> residual branch. The transformation still generates some weird noises
> >>> which maybe were removed by that filter. I don't know.
> >>>
> >>> No hurry but TODO: could you take a look at the CombFilter, fix and
> >>> reenable it and also check the overall transformation to check whether
> >>> it is sounding as it was intended to sound?
> >>>
> >>>
> >>> -----------------------------------------------------------------------
> >>>-
> >>>
> >>> _______________________________________________
> >>> Clam-devel mailing list
> >>> Clam-devel at llistes.projectes.lafarga.org
> >>> https://llistes.projectes.lafarga.org/cgi-bin/mailman/listinfo/clam-dev
> >>>el
> >>
> >> _______________________________________________
> >> Clam-devel mailing list
> >> Clam-devel at llistes.projectes.lafarga.org
> >> https://llistes.projectes.lafarga.org/cgi-bin/mailman/listinfo/clam-deve
> >>l



-- 
David García Garzón
(Work) dgarcia at iua dot upf anotherdot es
(Home) vokimon at telefonica adot net
http://www.iua.upf.edu/~dgarcia
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://lists.clam-project.org/pipermail/clam-devel-clam-project.org/attachments/20070415/7b2aaae7/attachment-0001.sig>