[asterisk-dev] Asterisk goes Spatial Conferencing: the STEAK project

[Dennis Guse]

Hello,

spatial audio for telephone conferencing has been proclaimed as a
silver bullet for speech-based telephone conferencing.
For research, we needed a _production-ready_ system that provides a
centralized conferencing bridge being able to provide binaural
synthesized spatial audio (i.e., for a pair of headphones).
As none was available (except for proprietary solutions), we chose to
get our hands dirty and extended Asterisk.
Most important, the modifications were required to  be complaint to
VoIP standards (no fancy protocol extensions).
We succeeded and the modifications are now at a stage that makes them
suited to be merged back upstream.
Or at least start the discussion on this topic.

Technical Details (at the moment the modifications are based upon 13.6.0):
* Enabled OPUS (with incoming stereo and outgoing stereo [interleaved])
* Extended softmix for stereo support (downmixing)
* Extended the default confbridge (basically added a convolution engine)

We chose OPUS (beside being used for WebRTC) as it allows to
RTP-channels with stereo (L16 or AMR-WB+ are not really alternatives).
Convolution was implemented via libfftw3 and the required HRTFs are at
the moment compiled into Asterisk.

We would now like to bring the changes upstream as we think this is an
important feature (that might set Asterisk apart) and is not too hard
to maintain.

More detailed information are available at our website:
http://steakconferening.de
and  in the source code (branch >>steak-13.6.0<<):
https://github.com/steakconferencing/asterisk

In addition, we host WebRTC- based demo (the real system):
https://demo.steakconferencing.de

Best regards,
---
Dennis Guse
TU Berlin
dennis.guse at alumni.tu-berlin.de