[vorbis] Graceful degradation of signal

[Johan Ovlinger]

Eric, big thanks for your answer. Let me just make sure I'm following
you. 

"Eric Scheirer" (Mon, 15 May 2000 18:23:58 EDT) proclaims:
> "Scalability" is the name that MPEG uses for the notion of having a
> bitstream that can be unpeeled.  That is, given a 128 kbps
> bitstream, you can easily extract from it a 96 kbps bitstream that
> has a minimal degradation in quality.

1) I assume that by "minimal" you are being precise(ish) in that the
   new stream sounds as good as a bitstream encoded directly to 96
   kbps could be expected to sound.

2) is the extraction reversible? Can I extract 128 kbps to 96kbps + 32
   kbps, and then later rebuild the 128 kbps stream from the 96 kbps
   and 32 kbps streams? Or is the extraction lossy in itself; so that
   there is no way to get the 32kbps differential stream?

> The way MPEG-4 does it, undoubtedly patent- pending, is to represent
> the stream as a "base layer" plus one or more "scalability layers."
> The layers are not each complete sound signals, but represent extra
> quantization--like coding the residual--of the original sound.

If we encode the residuals like that, will 96 kbps base + 32 kbps
residual sound as good as 128 kbps encoded directly? My intuition says
no, but it doesn't know what it's talking about.

> In MPEG-4 (the first MPEG audio standard to provide error
> resiliance), you pay about 25% penalty to get statistically
> indistinguishable quality from the uncorrupted signal at 1e-3 bit
> error rate.

At first glance that seems really high. I assume that's poisson
distribution? Is this a valid assumption to make on the internet?
Anyway, 25% is a 4+1 parity (or some less naive coding). I would have
thought that would correct much higher error rates. 

> I don't think MPEG has paid a lot of attention to error concealment,
> which is what you have to do when you have packet loss.  Private
> techniques (RealAudio) probably do this better.

Yes, from their promotional litterature, IIRC, they do something akin
to this, so that a dropped packet means a loss of quality for n
quanta, rather than silence for 1 quantum. This whole question started
as a gedanken experiment to understand how to acheive that.

Ta for the infos Erik

Johan

--- >8 ----
List archives:  http://www.xiph.org/archives/
Ogg project homepage: http://www.xiph.org/ogg/
To unsubscribe from this list, send a message to 'vorbis-request at xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is needed.
Unsubscribe messages sent to the list will be ignored/filtered.