On Tue, 9 May 2000, Mark Stier wrote:
> Hello,
>
> > File size is easy enough to measure, but "sound quality" is not.
> > It is purely subjective.
> > VERY HARD to define on a computer.
>
> Are there any examples demonstrating (so I can _hear_ them) that a
> 'mathematical' comparison algo that tries to keep differences at a minimum
> doesn't work?
>
> I believe it, but I wonder how bad it really is :-)
It's often easier to consider how easy it is to make 'mathmatically bad'
compression but still sound the same. This shows the same thing, that it's
hard to model the human perception of sound.
It's super easy in time-domain:
Take a input signal and multiply it by -1. Your RMS error is hugh, but it
sounds the same.... or take a signal, cut off the first 1ms of samples,
compair the results. They will sound the same. dc offsets,
amplification, compression and other non-linear effets, etc all
mathmaticaly look HORRIBLE, but don't affect the audio that much.
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
