[FFmpeg-devel] GSoC 2008 qualification task

[Tomasz Żernicki]

Dnia 29 marca 2008 19:28 Benjamin Larsson <banan at ludd.ltu.se> napisa?(a):

> Michael Niedermayer wrote:
> > On Fri, Mar 28, 2008 at 07:05:21PM +0100, Tomasz Z.ernicki wrote:
> >> Hello,
> >>   My name is Tomasz Z.ernicki. I would like to participate in the GSoC 2008. I have some DSP and audio coding background as well as experience in C programming.
> >>   According to the ?Idea List? I would like to choose ?AAC-LC Encoder? as my primary GSoC task.
> > 
> >>   On the other hand I want to choose ?Nellymoser forward transform? as my qualification task.
> > 
> > See libavcodec/nellymoserdec.c theres a imdct some reordering, windowing
> > and stuff :)
> > Goal is a clean inverse transform of this transform. It has to be fast (no
> > O(n^3) matrix multiply!) good news is, all building blocks should be already
> > written (we have mdct, fft, ... code)
> > I also think you should write some selftest code to ensure that your transform
> > works correctly that is your transform -> existing transform is (nearly) the
> > same as what was input at the start.
> > And i think benjamin will want to comment on this / correct me ...
> > 
> 
> It's correct. But you need to use the FFmpeg mdct function and then
> perform some shuffling so the samples will be decoded by the nellymoser
> decoder transform. It is basically the first step of a nellymoser encoder.

Dear All,
Currently I?m trying to figure out how does the transforms in Nellymoser Asao codec work. 
Due to forthcoming deadline of GSoC 2008, I would like to know what are your expectations
connected with "AAC-LC Encoder" task? 
What does exactly mean "flexible framework for a serious psychoacoustic implementation"?
What features would you like me to implement?

I would appreciate any help.

Greetings
Tomasz ?ernicki