On 08/06/2012 06:12, David Rowe wrote:
> Hello Patrick,
>
>> David, Jean-Marc, would it be possible to backport the methods of the
>> 1200 bit/sec version to 2500 bit/sec and get even better quality on 2500
>> bit/sec? I would say 2500 bit/sec is not yet toll quality, but maybe it
>> could be brought there.
> I tried some clever tricks but stumbled across a simple one that works
> surprisingly well - I just changed the frame rate from 20 to 40ms. So
> at 1400 bit/s the amplitude information (encoded by LSPs) is updated
> every 40ms rather than every 20ms. I still update the pitch and
> voicing every 20ms at the lower rates.
Increasing frame rate seems a sensible approach generally to trying to
reduce bit rates.
Have you considered:
- 30ms/40ms frames as the default
- predicting amplitude changes (across multiple frames?) and using the
bits saved for more LSPs (codebook/PPM/etc)
> At around 4000 bit/s we could start sending the phases of each harmonic,
> that would make a significant difference. Currently the phases are
> discarded, a we use a rule based approach to synthesise them at the
> decoder. This phase model is the cause of the "buzzy" artefact
> discussed a week ago by Ed.
It's super easy to suggest work for other people to do, but...
- If this would give a decent step-up in quality then I should think
it's worth implementing if only to set an upper bound on what can be
achieved at the lower rate?
- Use your trick to half the bits used for amplitude as above, and use
the saved bits for phase transmission, potentially getting more quality
into the 2.4kbit version?
- Have a long multi-frame length predictor looking to try and save bits
if either amplitude or lsp use can be better predicted? (is this a silly
idea?)
There aren't many codecs in the 4Kbit arena. Of the top of my head I
can think of the AMR-nb from the 3G telecomms standards which drops down
to 4.75kbit and has I think MOS around 3.2 ish (amazing). Some
instructions on compiling against asterisk here (beware patents and
ensure legal in your jurisdiction)
http://www.venturevoip.com/How-to-asterisk.1.8_SRTPTLS_snom300_pgsm.pdf
It would be very interesting to see a toll quality codec in the 4Kbits
region. This kind of bandwidth works for GPRS and the obvious
competition would be low rate iLBC. Any chance codec2 might compete in
that space?
Cheers
Ed W
------------------------------------------------------------------------------
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat landscape has changed and how IT managers can respond. Discussions
will include endpoint security, mobile security and the latest in malware
threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/
_______________________________________________
Freetel-codec2 mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/freetel-codec2
