i would prefer c or assembly implementation, but anything would do
in the spec they mention ITU-T's reference code, but i didn't find it.
Thanks,
Dg.
Full Version: g.723.1
does anyone knows where i can find source-code that implements this dual rate speech coder, g.723.1 ?
i would prefer c or assembly implementation, but anything would do
in the spec they mention ITU-T's reference code, but i didn't find it.
Thanks,
Dg.
i would prefer c or assembly implementation, but anything would do
in the spec they mention ITU-T's reference code, but i didn't find it.
Thanks,
Dg.
You can buy it from ITU-T online 
It is covered by patents, and when I was interested in it I couldn't find any freely available source - ITU reference code is slow, btw.
http://www.intel.com/software/products/ipp...p30/samples.htm claim to have g.723.1 source, but at my company we use IPP 2.0
Also, I've find the object code of g.723.1 encoder with some developer-free license, but I can't remember where
try google
It is covered by patents, and when I was interested in it I couldn't find any freely available source - ITU reference code is slow, btw.
http://www.intel.com/software/products/ipp...p30/samples.htm claim to have g.723.1 source, but at my company we use IPP 2.0
Also, I've find the object code of g.723.1 encoder with some developer-free license, but I can't remember where
try google
It appears only to be available on diskettes and CD, from ITU, not on paper.
i don't care about patents. my company will pay for it . i need sample sources 
.
itu's spec is incomplete cause it includes "pointers" to the reference implementation.
@ivan
what is covered in ipp ?
. i need sample sources .itu's spec is incomplete cause it includes "pointers" to the reference implementation.
@ivan
what is covered in ipp ?
QUOTE(Ivan Dimkovic @ Nov 16 2002 - 02:26 PM)
ITU reference code is slow, btw.
Also, I've find the object code of g.723.1 encoder with some developer-free license, but I can't remember where try google
Also, I've find the object code of g.723.1 encoder with some developer-free license, but I can't remember where
try google i'm planning to port this c code to assembly DSP anyway, so i don't care about the original implementation's speed.
and if IPP is implemented in assmebly x86, that's lots of work for me.
about google - no luck
.
You should buy ITU-T examlpe floating-point source code (or fixed-point, but I dunno if it exists for g723.1 - it do exists for g.729)
I think this is it:
http://www.itu.int/rec/recommendation.asp?...9611-I!AnnA
IPP won't help you to port the code to DSP because Intel offers x86 object-code only - hence there is no way to port it to anything else than IA-32 / IA-64
I think this is it:
http://www.itu.int/rec/recommendation.asp?...9611-I!AnnA
IPP won't help you to port the code to DSP because Intel offers x86 object-code only - hence there is no way to port it to anything else than IA-32 / IA-64
you can download it free of charge, even
QUOTE(Ivan Dimkovic @ Nov 16 2002 - 03:50 PM)
thanks ivan !!
any reason to prefer appendix a over appendix b ?
Annex A is fixed-point code,
Annex B is floating-point code
It all depends on your target hardware platform (is DSP floating point or fixed-point based), power consuption (integer decoder might consume less power even on floating point DSP-s), etc..
On IA-32 platforms (especially with SSE) floating point code might run faster than 32-bit fixed points.
Annex B is floating-point code
It all depends on your target hardware platform (is DSP floating point or fixed-point based), power consuption (integer decoder might consume less power even on floating point DSP-s), etc..
On IA-32 platforms (especially with SSE) floating point code might run faster than 32-bit fixed points.
yea, i've noticed it, and since my target is fixed point (16 or 24 bit) i only downloaded app A. it's just that i've noticed that Appendix A can be stored in 3 diskettes while App B requires 14 diskettes, so i thought it might include more goodies 
btw, you've mentioned this code's speed. what do you know about this code's quality ?
btw, you've mentioned this code's speed. what do you know about this code's quality ?
Hmm.. it's been a while since I played with it .. please don't take my word for it - only thing you could do is to compare it with some commercial g.723.1 codec. From experience, I could tell that reference ITU-T implementations are good enough for real usage, other than explanations.
Intel IPP offers optimized example of the ITU source code, so it is worth looking just because many DSP-s also offer optimized DSP libraries (fast convolutions, matrix operations, FFT, DCT, ...) - and you can easily identify which parts of the code could be optimized with replacing functions.
Intel IPP offers optimized example of the ITU source code, so it is worth looking just because many DSP-s also offer optimized DSP libraries (fast convolutions, matrix operations, FFT, DCT, ...) - and you can easily identify which parts of the code could be optimized with replacing functions.
QUOTE(Ivan Dimkovic @ Nov 16 2002 - 06:11 PM)
Intel IPP offers optimized example of the ITU source code, so it is worth looking just because many DSP-s also offer optimized DSP libraries (fast convolutions, matrix operations, FFT, DCT, ...) - and you can easily identify which parts of the code could be optimized with replacing functions.
optimizing code to DSP assembly is my forte (that's how i got my nickname
), but i need to know the code's quality is decent before i start porting.
Well, you can try to organize small listening test, or try to evaluate the output decoded results with PSQM tool (www.opticom.de)
Speech codecs usually have decent free reference implementations (even MPEG-4 CELP and PAR narrowband codecs are very nicely implemented, compared to more than awful AAC encoder implementation)
Speech codecs usually have decent free reference implementations (even MPEG-4 CELP and PAR narrowband codecs are very nicely implemented, compared to more than awful AAC encoder implementation)
thanks again, ivan.
now i need to create myself a small testing framework..
i browsed rfc2361 and found :
i know that vivo codec is g.723 @ 8kbps, but don't know anything about the other two.
i bet lucent's g.723 entry is the one i'm looking for (5.3/6.4).
does any1 know where i can find a codec with registration of 0x0059 ?
i guess i can build a winamp plugin out of itu's reference code, but it would be nicer to find a working codec and focusing on the dsp programming.
now i need to create myself a small testing framework..
i browsed rfc2361 and found :
QUOTE
A.49 Lucent G.723
WAVE form Registration Number (hex): 0x0059
Codec ID in the IANA Namespace: audio/vnd.wave;codec=59
WAVE form wFormatTag ID: WAVE_FORMAT_LUCENT_G723
Contact:
Ray Jones
(raykj@lucent.com)
Lucent Technologies
A.85 Vivo G.723
WAVE form Registration Number (hex): 0x00111
Codec ID in the IANA Namespace: audio/vnd.wave;codec=111
WAVE form wFormatTag ID: WAVE_FORMAT_VIVO_G723
Contact:
Vivo Software
411 Waverley Oaks Road, Suite 313
Waltham, MA 02154 USA
(617) 899-8900
A.87 Digital G.723
WAVE form Registration Number (hex): 0x0123
Codec ID in the IANA Namespace: audio/vnd.wave;codec=123
WAVE form wFormatTag ID: WAVE_FORMAT_DIGITAL_G723
Contact:
John Forecast
forecast@shell.lkg.dec.com
508-486-5264
Digital Equipment Corporation
146 Main Street
Maynard, MA 01754-2571 USA
1-800-DIGITAL
WAVE form Registration Number (hex): 0x0059
Codec ID in the IANA Namespace: audio/vnd.wave;codec=59
WAVE form wFormatTag ID: WAVE_FORMAT_LUCENT_G723
Contact:
Ray Jones
(raykj@lucent.com)
Lucent Technologies
A.85 Vivo G.723
WAVE form Registration Number (hex): 0x00111
Codec ID in the IANA Namespace: audio/vnd.wave;codec=111
WAVE form wFormatTag ID: WAVE_FORMAT_VIVO_G723
Contact:
Vivo Software
411 Waverley Oaks Road, Suite 313
Waltham, MA 02154 USA
(617) 899-8900
A.87 Digital G.723
WAVE form Registration Number (hex): 0x0123
Codec ID in the IANA Namespace: audio/vnd.wave;codec=123
WAVE form wFormatTag ID: WAVE_FORMAT_DIGITAL_G723
Contact:
John Forecast
forecast@shell.lkg.dec.com
508-486-5264
Digital Equipment Corporation
146 Main Street
Maynard, MA 01754-2571 USA
1-800-DIGITAL
i know that vivo codec is g.723 @ 8kbps, but don't know anything about the other two.
i bet lucent's g.723 entry is the one i'm looking for (5.3/6.4).
does any1 know where i can find a codec with registration of 0x0059 ?
i guess i can build a winamp plugin out of itu's reference code, but it would be nicer to find a working codec and focusing on the dsp programming.
This is a "lo-fi" version of our main content. To view the full version with more information, formatting and images, please click here.