From: <mr.escape_at_gmx.de>
Date: Sun, 21 Dec 2003 19:29:51 +0100

At 18:59 21.12.2003, you wrote:
>Björn Stenberg wrote:
>
> > RTFA. At least the C'T test *was* double-blind. The testers downloaded
> > seven .wav files and rated them without knowing which file was which
> > codec.
>
>What you describe is a blind test. For it to be double-blind, the persons
>administering the tests would also need to be unaware of which file each
>tester got until after the test was completed. It was unclear whether they
>were or were not. A blind test is certainly of more value than one which is
>not, but it is not as good as a double-blind test.

Each participant got a set of wav files (using some lossless compression to
reduce download size). Each participant got the very same files but with a
personal set of filenames so cheating was not that easy. One of the files
was the original and the other files were the decompressed version of the
output of the different codecs. The original file was not marked. The task
was to rate/sort the files by quality.


>You are also extrapolating too much from that test. While Ogg Vorbis might
>have been superior at the low bit rates they tested (64, 128, and 160), it
>may not have fared nearly so well in a test against a high quality VBR MP3.
>We just don't know.

How god is better than the original. If a low bitrate version of a file is
indistinguishable from the original (except by "acoustically challenged"
people), why do we need more bits or what is the benefit of artificially
enlarge the compressed file?

mr.escape
Received on 2003-12-21