[FFmpeg-devel] [RFC] AAC Encoder

Gabriel Bouvigne bouvigne
Wed Aug 20 16:43:09 CEST 2008

Michael Niedermayer a ?crit :
> What i dislike on the highpass method is that it is a heuristic checking for
> some pattern very commonly occuring together with cases that are better coded
> with short windows. It does not check which of the transform sizes is better
> at decorrelating the data which is what actually matters.
> So no, i do not belive at all that the highpass method is a good choice :)

I understand your point, and in the way you view things, you are right: 
the highpass method is "just" an heuristic, which dates back from the 
times when audio encoders were far from real-time.
Usually there are things more important quality-wise than replacing an 
highpass short blocks triger method with a more optimal one, but on the 
positive side you can play with alternative short blocks triger methods 
without the need to read a book on psychoacoustics first.

I guess that a frequency based method such as this would please you more:

*compute the 9 short time->freq transforms
*compute power per scalefactor band
*check correlation of the sfb powers across subsequent short blocks
*potentially trigger a switch to short blocks based on a low correlation

(obvious downside is that this requires to compute the 9 short 
transforms, even in cases where you will use a long block)

Gabriel Bouvigne
personal page: http://gabriel.mp3-tech.org

More information about the ffmpeg-devel mailing list