Gabriel Bouvigne wrote:
> So maybe we could imagine to start allocating bits starting from the middle
> subband (according to ear sensitivity), like about 1500-2000 Hz, and then
> allocating the just lower subband, then the just higher,...
>
>
>
> What do the people on this list think about such a bit allocation process
> starting from the middle subband?
When I hear errors on all coders around the 96kbit/s mark, I estimate that they
are between 1kHz and 3.5kHz, and (though I know next-to-nothing about
psychoacoustics) I'd be surprised if these bands weren't treated as `critical'
in terms of noise and quantizing bits anyway.
For example, the RealAudioG2 (`cook') coder down at 20kbit/s seems very
accurate in the < 3.5kHz area, but seems to call for terrible amounts of
filtered noise to be added at the top end (4kHz +), to surround the sounds it
*does* encode, maybe to disguise poor coding up there? Try listening to some
classical piano music through it in mono, and you'll see what I mean...
Another thought: PRE-EMPHASIS AT LOW BITRATES... I was running LAME down at
16kbit/s, and found that, no matter how I changed the lowpass filter, you've
selected some great default settings. However, I was able to improve the
*perceived* quality by emphasising the treble above 3.5kHz on replay. Can this
sort of pre-emphasis be done, as an option, in the encoding easily, or am I
fooling my hearing?
John Hayward-Warburton
[EMAIL PROTECTED]
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
