There is no functionality like this in the code.
Downsampling in the
frequency domain will be difficult because of the MDCT and
probably
wouldn't reduce overall complexity much. But feel free to
investigate. Depending on the platform, it might be possible
to save
enough cycles through conventional optimizations to run the
standard
decoder and a good resampler instead.
Another option would be to zero out (not decode) the upper
MDCT
coefficients, do the standard inverse transform, then run a
fast,
low-quality downsampler. This would also allow you to do
less work in
dequantization, TNS, etc. since unused coefficients could be
skipped.
But this damages audio quality significantly, and it would
be best to
avoid this if possible.
At 02:41 PM 10/31/2007, syl wrote:
>Hi,
>
>Has anyone here worked on partially decoding AAC-LC
format? The
>purpose is to save CPU time and to get a lower decoded
data rate. The
>most straightforward approach, I guess, is to synthesize
only the
>lower bands. Can this be done by using half sized fft?
What are the
>others changes needed? Is the code for doing this in the
fixed point
>decoder already available?
>
>Thank you very much
>
>_______________________________________________
>Datatype-dev mailing list
>Datatype-dev helixcommunity.org
>http://lists.helixcommunity.org/mailman/listinfo/da
tatype-dev
_______________________________________________
Datatype-dev mailing list
Datatype-devhelixcommunity.org
http://lists.helixcommunity.org/mailman/listinfo/da
tatype-dev
|
