[FFmpeg-devel] [PATCH] NellyMoser audio decoder v2
Tue Oct 2 20:43:59 CEST 2007
On Tue, Oct 02, 2007 at 10:44:46AM +0200, Benjamin Larsson wrote:
> Michael Niedermayer wrote:
> > Hi
> > On Mon, Oct 01, 2007 at 02:25:59PM +0200, Benjamin Larsson wrote:
> >> Benjamin Larsson wrote:
> >>> [...]
> >>> Hi, I had some time yesterday and looked closer at the code. While there
> >>> is no doubt that the code is a TDAC implementation it doesn't seem like
> >>> the code can be replaced by the mdct code in ffmpeg. So I guess that the
> >>> fft replacement is what can be clean up for now. I'll sort out the table
> >>> generation for the rotations and the window.
> > if its not IMDCT then what is it?
> In the flash specs that are available they write it is just plain
> Fourier transform. And from reading the code they use a 50% overlap and
> the overlap window is a sine window (same as in most mdct based codecs).
> But the antialiasing stuff suggest that there is some other thing also.
> It looks a bit like the qdm2 transformation.
i just called it antialiassing because of where it was done, its entirely
possible that its something else ...
anyway, this is not hard to figure out, someone just has to look at the
output of the whole transform for a few test signals like
and if the output are pure sine/cosine (likely) then simplify it by finding
phase/frequency for each
from that it should be obvious which variant of dct/dst/dft it is
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
I do not agree with what you have to say, but I'll defend to the death your
right to say it. -- Voltaire
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: Digital signature
More information about the ffmpeg-devel