>
> > -X subtract 5db from all masking thresholds. The masking thresholds
> > are often too aggressive, causing problems like audible distortion
> > even with -V 0. Subtracting 5db from the ISO psy model is actually
> > recommended in "Application of a physiological ear model..."
> > by Frank Baumgarte (page 6):
> >
> http://www.tnt.uni-hannover.de/project/coding/audio/perception/publications.
> html
>
>
> I didn't understood this paper like that. It seems to me that the author
> never recommended to remove 5dB, but only done it to match the bitrate of
> its ear model.
> However, it's right that the ISO masking threshold are sometimes too
> aggressive, and sometimes not. So it seems a good thing, but 5dB is perhaps
> not the right value.
>
Yes, I wouldn't have understood the paper in that way either, except
for the fact that I got email from the auther, Frank Baumgarte. I had
told him that I thought the ISO model was a little too aggressive, and
he wrote:
> Mark,
> in my experiments a masked threshold reduction of 5 dB in all scalefactor
> bands leads to reasonable results. These results will be published and are
> now available from
>
> http://www.tnt.uni-hannover.de/project/coding/audio/perception/aes_17.html
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
