png-like activity for voice coder ?

This sounds a lot like the reasoning that brought us PNG ...
the question is should W3C help assembling the altruists out there to
do it ? e.g. by saying that we support this initiative by VON ?

I'm sure they'd love it - and it would be good for W3C as well -
but I'm not sure whether our members like it. Anyway, it's moratorium
time.

Anyway, it was/is on my activity page for a long time, and there
was interest to do this withing W3C by the AC representative of
Lucent.

For activity page, see

"A common fallback format for audio and video data"

http://www.w3.org/pub/WWW/Consortium/Prospectus/RealTime.html#anchor755890


------- Forwarded Message


From: Jeff Pulver <jeff@Pulver.COM>
To: VON.Mailing.List@enterprise.pulver.com
Subject: [VON]: Patents and Speech Coders 
Sender: owner-von@Pulver.COM
Precedence: bulk
Content-Type: text
Content-Length: 4513

Hi There,

In response to my recent posting regarding the development of a NextGen 
Codec which could freely be used by the VON industry - I received the 
following e-mail which I thought some of you might find interesting:

- -------------------

From: Rich Cox
To: jeff@pulver.com
Subject: Patents and Speech Coders

Jeff,
	let me introduce myself.  My name is Rich Cox and I now work for
AT&T Labs Research.  What I'm writing here represents my own personal
viewpoint and is not intended to be a representation of any official 
AT&T policy.  

	From 1993 - 1995 I was the editor responsible for
creating G.723.1 in the ITU.  It was not an easy task.  In fact, creating
speech coding standards in any body is not an easy task.  You would not
have wanted to pay for my travel bill, let alone my time.  Now multiply
the costs of my time and travel by 20 and you begin to get an idea
of how much it costs to make an ITU speech coding standard.  We pioneered
some new ground in speech coding standards in the ITU with G.723.1.
It was the first ITU speech coder to be specified in C code.  Thanks
to Intel the interoperable floating point C code was created based on
the bit exact fixed point C code.  It is the first ITU coder to specify
a silence compression scheme (voice activity detector plus comfort
noise generation).  Thanks to the work of the committee that did this,
all of it is specified in C code.  Other ITU standards have and will
follow these precedents.

	The world does not run on altruism, although one might wish it
so.  G.723.1 is based on fundamental work that goes back as much as 15 years.
The coder includes intellectual property from AT&T (part of which is 
now owned by Lucent), NTT, University of Sherbrooke, France Telecom/CNET, 
AudioCodes, Ltd., and DSP Group.  DSP Group collects the licensing fees 
representing its own share as well as those of AudioCodes, FT/CNET, and 
Sherbrooke.  To my knowledge, they are not collecting fees for AT&T, Lucent 
or NTT.  All of these companies and the university are entitled to
compensation for their investment in speech coding research, because
they hold patents that read on this standard.  That is why they can 
collect royalties.

	The ITU has created a second coder that meets most of the
requirements listed in your article.  Annex A of G.729 is low complexity
(about 10 MIPS on a DSP, which is 30% less than G.723.1), has a bit rate
of 7.9 kb/s, has a 10 ms frame size, and is designed to withstand
packet losses.  Its subjective quality is at least as good as G.723.1
and it has been tested in English and Japanese.  It is derived from
the original G.729 coder that was much more extensively tested than G.723.1.
Its one shortcoming, if you will, is that it, too, requires royalty
payments.  DSP Group and AudioCodes are not involved in it, but all the
others listed above are.  

	If someone else set out to create another speech coder, they 
would probably run up against the same patent holders once more.  For 
example, two new cellular speech coding standards were created for 
North America and Europe by Nokia in 1995.  Both of them use patents 
held by AT&T, NTT, and University of Sherbrooke.  These days 
there are a number of small companies that are entering the speech 
coding business.  The keyword is business.  Just as DSP Group and 
AudioCodes expect to collect royalties, so will other new companies.
And more established companies are going to want to continue as well.
If some group makes a coder as a de facto standard and it uses anyone
else's patents, the the patent owners will still be entitled to collect 
royalties or could even block usage of the coders.

	From my vantage point, I'd say stick with the standards.  They
are better tested and have undergone more scrutiny to discover any
items that might cause problems later.  That includes scrutiny
for intellectual property that could block the standard.  For Internet 
Telephony to really take off in a big way, there will need to be standards 
so that anyone can talk to anyone else.  We've come a long way since the 
Vocaltec phone in 1995 to where we are now with widespread support for 
H.323-based phones.  I expect that further progress will be made as well.
But don't expect companies that have invested millions of dollars into
technology and have gone to the trouble of patenting it to protect
their investments to now give it away on the Internet for free.   The
Internet may be free, but the intellectual property on it is not free.

			Rich Cox

------- End of Forwarded Message

Received on Friday, 8 November 1996 09:45:20 UTC