[vorbis] RC3?

Greg Wooledge greg at wooledge.org
Fri Jan 4 15:58:13 PST 2002


Brendan Ragan (reaper at post1.com) wrote:

> GW> What we need are real statistics.  Then we can say things like

> yeah, that's a really good idea. Couldn't we just burn .wav's to CDs
> and then get people to listen to them? (or do we want "purer"
> listening tests (i.e. abx and stuff)

That's fine if you have a bunch of people in one physical location (e.g.
a college campus).  Otherwise, you're going to have to ship CDs all over
the place.  That costs money.

Ideally we'd want some sort of ABX test, but that may be asking too much.

If you haven't done so yet, take a look at ff123's listening test which
involved a CVS snapshot of Vorbis from a couple months ago.  As far as I
know the test is actually ongoing, but I don't think anyone's submitted
any results lately.  <http://ff123.net/128test/instruct.html>

> Are we aiming these percentile ratings at the average user,
> audiophile or somewhere in between?

If you aim for the average user, nearly every -q rating is going to be
100% transparent.  The average user can't hear artifacts because she
doesn't have enough experience in listening tests.  I've been trying to
hear artifacts in Ogg for months now, and I'm still overwhelmed by my
own inadequacy.

If you aim for the audiophile, there aren't enough of them available.
ff123 only has about two dozen listeners on his test -- after 2-3 months.

> GW> Of course, this would require a massive set of data (listening
> tests
> GW> with real hardware), and therefore time.

> True, but it would only have to be done before 1.0 - yes? (which
> should give us at least 2 - 3 months..)

(No comment.)


-- 
Greg Wooledge                  |   "Truth belongs to everybody."
greg at wooledge.org              |    - The Red Hot Chili Peppers
http://wooledge.org/~greg/     |


-------------- next part --------------
A non-text attachment was scrubbed...
Name: part
Type: application/pgp-signature
Size: 241 bytes
Desc: not available
Url : http://lists.xiph.org/pipermail/vorbis/attachments/20020104/69044f51/part.pgp


More information about the Vorbis mailing list