[music-dsp] IMDCT filterbank

[Joshua Scholar]

Ooh.  Think of it this way:

If it was a normal DCT then a single band can't even HAVE any locality no
matter what the input impulse was - it just has a sinewave that's block
windowed) varying response.

Since a MDCT overlaps by one half (that's what we're talking about right?)
then the response to an impulse is the sum of two overlapping windowed sine
waves.  That's not much locality.


Joshua Scholar

----- Original Message ----- 
From: "Anup KC" <anupkc at emuzed.com>
To: "a list for musical digital signal processing"
<music-dsp at shoko.calarts.edu>
Sent: Wednesday, February 02, 2005 9:59 PM
Subject: Re: [music-dsp] IMDCT filterbank


> Hi Joshua
>
> Thank you very much for the reply..
>
> The explanation you provided gives a very good picture of an ordinary
> transform ( say DCT or DFT) seen as a filterbank... but what is the impact
> of lapping ( IMDCT being a lapped transform)..how does it affect the
> features of the filterbank
>
> - Anup
> ----- Original Message -----
> From: "Joshua Scholar" <joshscholar at yahoo.com>
> To: "a list for musical digital signal processing"
> <music-dsp at shoko.calarts.edu>
> Sent: Thursday, February 03, 2005 11:15 AM
> Subject: Re: [music-dsp] IMDCT filterbank
>
>
> >
> > ----- Original Message -----
> > From: "Joshua Scholar" <joshscholar at yahoo.com>
> > To: "a list for musical digital signal processing"
> > <music-dsp at ceait.calarts.edu>
> > Sent: Wednesday, February 02, 2005 9:41 PM
> > Subject: Re: [music-dsp] IMDCT filterbank
> >
> >
> > >
> > > Because it's downsampled, I think that if you were to filter out all
but
> > one
> > > band, you'd find a different impulse response for each position in the
> > > block, but still somewhat localized in frequency.
> > >
> > > You would find similar, but less extreme effects with other
downsampled
> > > filterbanks such as quadrature mirror filterbanks.
> >
> > By the way, in a QM filter, if the (arbitrary) filter it's based on is a
> > perfect one, then that dependence on position must disappear, because
> > perfect is perfect.
> >
> > I have no idea if there's some class of non-perfect filters that would
> also
> > get rid of that dependence.
> >
> > Joshua Scholar
> >
> > --
> > dupswapdrop -- the music-dsp mailing list and website:
> > subscription info, FAQ, source code archive, list archive, book reviews,
> dsp links
> > http://shoko.calarts.edu/musicdsp
> > http://ceait.calarts.edu/mailman/listinfo/music-dsp
>
> --
> dupswapdrop -- the music-dsp mailing list and website:
> subscription info, FAQ, source code archive, list archive, book reviews,
dsp links
> http://shoko.calarts.edu/musicdsp
> http://ceait.calarts.edu/mailman/listinfo/music-dsp