[mirtoolbox] Re: singing voice timbre
- From: Pietro Guccione <pietro.guccione@xxxxxxxxx>
- To: mirtoolbox@xxxxxxxxxxxxx
- Date: Fri, 22 Feb 2013 09:04:26 +0100
Dear all
I'm starting to perform a study on the effects of codec's (as MP3) on
features of speech (mainly pitch).
Even if MP3 saves psychoacoustic effects of sound/speech, none can be said
about extracted features as pitch, energy level, formants and so on.
It depends, clearly, on the level of bitrate required to the compression.
Anyone knows if similar studies are already available or how to approach to
such problem?
Thanks in advance
PG
On Thu, Feb 21, 2013 at 12:59 PM, Olivier Lartillot <olartillot@xxxxxxxxx>wrote:
> Dear Pauline,
>
> In current version of mirtoolbox, you can extract:
> - the energy envelope, so that you can detect phonemes, segment
> automatically, and assess the dynamics and the attack/release
> characteristics for instance.
> - the f0 curve
> - timbral characteristics, if they could help (brightness, roughness),
> - many other features could be useful as well? For instance consonant
> sounds could be detected and characterized in some ways..
>
> I am currently planning to add also:
> - tracking of the harmonics (f1, f2, etc.), so that we could also assess
> the harmonic/noise ratio,
> - estimation of formants.
> I might have something ready for the summer, but I can't promise..
>
> Cheers,
>
> Olivier
>
> On Feb 19, 2013, at 3:36 PM, pmouawad@xxxxxxxx wrote:
>
> >
> > Dear List,
> >
> > Anyone knows whether the mirtoolbox extracts features of singing voice
> timbre?
> >
> > Thanks,
> > Cheers,
> > Pauline
> >
>
>
>
Other related posts:
