[AG-TECH] RAT audio quality - Alternative codecs?

[Piers O'Hanlon]

2009/9/22 Henry Hernando Caballero Corzo <hh.caballero921 at uniandes.edu.co>:
> Hi everyone,
> Hope you can advise me in the following issue.
> Scenario:
> Say we're holding a videoconference which joins people from two different
> types of nodes:
> -A conference room with an adequate AG node setup (audio equipment for echo
> cancelling (ClearOne RAV 600), ambient microphone, high-end speakers, etc.)
> -A "casual participant" (an academic who joins the conference from his
> workplace and uses his personal AG-equipped laptop sporting integrated
> microphone and speakers)
> Then a lot of -predictible- problems related to audio arise. Audio quality
> can degrade so badly (enormous echo and noise) that ends up ruining the
> meeting. We certainly know we're dealing with a "non-recommendable"
> scenario, but the fact is, we need it to work.
> First question:
> 1. How do you guys deal with this kind of setups? (if there's some way)
> Going deeper in this issue, I -almost inevitably- turned my eyes on RAT.
> Even though RAT works well under controlled setups, We certainly expect it
> to perform much better when used in uncontrolled environments like the
>
RAT does have some echo supression which you can try out (under the
"Audio" settings).
Additionally we regularly use echo cancelling speaker boxes which can
be found quite cheaply now. They range from the well known Phoenix
Audio ones (http://www.phnxaudio.com/) to the more generic ones. Also
for personal set ups we generally recommend headphone and mic use.

> described above. So, there're some more questions to ask:
> 3. Is RAT lacking important features like: ultra-wideband sampling, echo
> cancelling, noise suppression, active voice detection, discontinuous
> transmission (among others) ?
>
RAT has the L16 codec which is wideband uncompressed audio, and the
WBS codec. We would welcome any further integration of public domain
wideband codecs (G722 is a good candidate). RAT has echo suppression
(as echo cancellation whilst possible is still quite onerous in
software). It also has silence suppression which results in
discontinuous transmission in periods of silence.

Active Voice detection is largely a feature of the codec in use - and
RAT's GSM codec performs some VAD. RAT doesn't provide for noise
supression, though this can be done in some sound cards and OSes (OSX
offers it as an option - though the OSX audio for RAT is still not
very good - we do hope to fix this).

> 4. Or it's all about tweaking with RAT options? (some of the above features
> seem to be included, not sure if they work well though)
> 5. What about Speex[1] and/or SILK[2] codecs? Would it be feasible to
> integrate these codecs to the Access Grid?
>
It's certainly possible to implement these codecs into RAT but we
don't have the resources to do it - we'd welcome anyone else to the
integration. Whilst NCHC did purportedly implement Speex into RAT we
never gained access to the source code for it.

> * I'm aware of this development[3], which is part of a contribution from
> NCHC. It involves the integration of Speex into RAT, but I'm really not sure
> if these features are included in the current version of RAT that ships with
> the latest AG release
> * SILK claims excellent voicecall quality and the truth is, Skype can in
> fact deal with our "problematic setup" with absolutely no problem

Skype is certainly an option for small scale meetings - Whilst AG is
more oriented for multi-site and multi-camera room deployments.

Piers.

> Thanks in advance!
> Henry Caballero
> ----------------
> AG node administrator - Universidad de los Andes
> (http://ag-mox.uniandes.edu.co)
> [1] http://speex.org/
> [2] https://developer.skype.com/silk
> [3] http://www.nchc.org.tw/en/core_content/download.php?FILE_ID=152