Speex dev - Aug 2004 - --dtx alone does nothing?

I'm running a:
for band in n w u
  do
    for quality in 0 1 2 3 4 5 6 7 8 9 10
      do
        for complexity in 3
          do
            for vad in "" "--vad"
              do
                for dtx in "" "--dtx"
                  do
                    echo speexenc -${band} --quality ${quality} --comp
${complexity} ${vad} ${dtx} input.${band}.wav
output.${band}.${quality}.${complexity}.${vad%%--}.${dtx%%--}.speex
                    /usr/bin/time speexenc -${band} --quality ${quality} --comp
${complexity} ${vad} ${dtx} input.${band}.wav
output.${band}.${quality}.${complexity}.${vad%%--}.${dtx%%--}.speex
                  done
              done
          done
      done
  done

on a set of .wav files generated from ripping a CD (Doug Stanhope/Die Laughing)
which is all talk with some applause (sorry Doug).  I sox'd them down to 8k,
16k, and 32k so that speex would have something good to chomp on.

So far, I've noticed that --dtx alone doesn't seem to do anything, but
--vad does
and the combination --dtx with --vad does more.  The lone exception is for wide
and ultra-wide coding at quality 0, --vad, alone, makes the file slightly
larger.
At all other sized the ordering is always: plain==--dtx, then --vad, then --vad
with --dtx.

Does that makes sense?  I would have thought that --dtx alone would have doen
it,
but maybe it needs perfect digital silence to work?  Or only uses the voice
activation
to know when there's silence?

Another point I'd just like clarified is how to specify bit rates.  There
seem to be
four different ways to do it:

1) --quality
2) --bitrate  (CBR?)
3) --bitrate with --vbr
4) --abr

Is that right?  Can someone point me to a description of what to expect from
each
of them?  Any chance of geting a histogram output like LAME has for it's
variable
bit rate encoding?  Heck, I'd be willing to code it. :)

Oh, please, no comments on my shell script. :)  Yes, I mean ##, not %%, but you
try to remember the difference at 2am. ;)

Cheers,
David

<p>--- >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
'speex-dev-request@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.

> So far, I've noticed that --dtx alone doesn't seem to do anything,
but --vad does
> and the combination --dtx with --vad does more.  The lone exception is for
wide
> and ultra-wide coding at quality 0, --vad, alone, makes the file slightly
larger.
> At all other sized the ordering is always: plain==--dtx, then --vad, then
--vad
> with --dtx.
All you say look normal to me.
> Does that makes sense?  I would have thought that --dtx alone would have
doen it,
> but maybe it needs perfect digital silence to work?  Or only uses the voice
activation
> to know when there's silence?
DTX requires the VAD to know when there's silence, so it can't do
anything without --vad. I'll add a warning for that. 
> Another point I'd just like clarified is how to specify bit rates. 
There seem to be
> four different ways to do it:
> 
> 1) --quality
Note that in narrowband mode, there are less than 11 different modes, so
quality 1 and 2 are the same, the same for 3 and 4, 5 and 6, and 7 and
8.
> 2) --bitrate  (CBR?)
yes, --bitrate is for CBR. Note there are only a few bit-rates available
so the one chosen is the first mode with a bit-rate less or equal the
the bit-rate specified.
> 3) --bitrate with --vbr
With --vbr, you control the bit-rate with --quality (which is a float in
vbr mode, but an int otherwise).
> 4) --abr
This is how can specify a bit-rate in vbr mode (no need to have --vbr
too).
> Is that right?  Can someone point me to a description of what to expect
from each
> of them?  Any chance of geting a histogram output like LAME has for
it's variable
> bit rate encoding?  Heck, I'd be willing to code it. :)
That would be welcome. There could be a --histo switch or something like
that.

        Jean-Marc


-- 
Jean-Marc Valin, M.Sc.A.
LABORIUS (http://www.gel.usherb.ca/laborius)
Université de Sherbrooke, Québec, Canada

<p>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 242 bytes
Desc: signature.asc
Url :
http://lists.xiph.org/pipermail/speex-dev/attachments/20030114/6cf57280/signature-0001.pgp

> All you say look normal to me.
Okay, clarification requires less code to be written than a bug. :)
> > Does that makes sense?  I would have thought that --dtx alone would
have > doen it,
> > but maybe it needs perfect digital silence to work?  Or only uses the
voi> ce activation
> > to know when there's silence?
> 
> DTX requires the VAD to know when there's silence, so it can't do
> anything without --vad. I'll add a warning for that.=20
Okay, my second guess was right.  I'm not completely insane today.
> > Another point I'd just like clarified is how to specify bit rates.
There>  seem to be=20
> > four different ways to do it:
> >=20
> > 1) --quality
> 
> Note that in narrowband mode, there are less than 11 different modes, so
> quality 1 and 2 are the same, the same for 3 and 4, 5 and 6, and 7 and
> 8.
Okay, I probably would have noticed that if I had stared at the data for too
long.
> > 2) --bitrate  (CBR?)
> 
> yes, --bitrate is for CBR. Note there are only a few bit-rates available
> so the one chosen is the first mode with a bit-rate less or equal the
> the bit-rate specified.
Okay, so just --bitrate gives CBR.
> > 3) --bitrate with --vbr
> 
> With --vbr, you control the bit-rate with --quality (which is a float in
> vbr mode, but an int otherwise).
Okay, so for VBR, one sets a bit rate with quality which is a float.
> > 4) --abr
> 
> This is how can specify a bit-rate in vbr mode (no need to have --vbr
> too).
So for VBR you can use --abr as well to specify an average bit rate.

So there should be some equivalency between --abr values and --quality
values?  Or do they effect the rate in different ways.
> > Is that right?  Can someone point me to a description of what to
expect f> rom each
> > of them?  Any chance of geting a histogram output like LAME has for
it's > variable
> > bit rate encoding?  Heck, I'd be willing to code it. :)
> 
> That would be welcome. There could be a --histo switch or something like
> that.
Okay, I'll start on it. :)

I'm still confused as to all of the different ways to specify the bit rate/
quality of an encoding.  Could you enumerate the valid combinations of
switches?

Cheers,
David
--- >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
'speex-dev-request@xiph.org'
containing only the word 'unsubscribe' in the body.  No subject is
needed.
Unsubscribe messages sent to the list will be ignored/filtered.