On Thu, 18 Nov 1999, Robert Hegemann wrote:
> OK, here is finaly a patch, that solves the
> problems not only with "Glockenspiel" gspi35_1.wav
> but with "Glockenspiel" gspi35_2.wav too.
>
> If a scalefactor band contains a peak Takehiro described
> earlier, then the average distortion within such a
> band will be much lower than the maximum distortion.
> By now Lame ignores such peaks, as distortion
> is defined as the average distortion over a scalefactor
> band only, and the peak becomes audible.
> With this patch Lame will define the distortion in such
> a band as something inbetween the average and the maximum,
> depending on the distance of them.
I had a chance to do some testing.
All I can say is.. WOW. This really improves my tests of VBR stupidity. :)
It does encode at a higher bitrate, but consistnatly sounds a lot better..
With gspi35_1.wav and -v -V0 it's 105Kb/s vs 87 Kb/s but it sounds SO much
better.
Has anyone ever considered tossing the stupid bands as far as
psycoacoustics are concerned, it seems much more sensible to do all
calculations over the non-segmented frequence space and breakup into bands
just for packing into the file. This would probably also make more
sophicated masking (like temporal masking) easier to impliment.
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
