Mathematicians wanted (was RE: [Asterisk-Users] Best echo canceller?)

[Kris Boutilier]

> -----Original Message-----
> From: asterisk-users-bounces at lists.digium.com
> [mailto:asterisk-users-bounces at lists.digium.com]On Behalf Of Andrew
> Kohlsmith
> Sent: Thursday, September 29, 2005 2:23 PM
> To: asterisk-users at lists.digium.com
> Subject: Re: [Asterisk-Users] Best echo canceller?
> 
> 
> On Thursday 29 September 2005 17:04, Claudio Canseco wrote:
> >  In your experience what is the best choice for echo canceller ?
> > Which one should work better: STEVE, STEVE2, MARK, MARK2, MARK3, KB1 ?
> 
> KB1 is a refactored MARK2 which seems to work VERY, very well.
> 
{clip}

The refactoring applied to MARK2 to create KB1 was basically intended to make the code generally consistent with the Texas Instruments whitepaper referenced in the comments at the top of the file. In that document they completely outline the operating theory of one particular echo cancellation algorithm and completely document an implementation of it in a general purpose TI processor. The implementation is also benchmarked and deviations/performance issues explored. The MARK2/KB1 implementation is not a 100% complete version of the reference code - there are some autotuning elements and perhaps additional optimizations suggested by TI that can certainly still be implemented.

I would strongly encourage anyone with a good understanding of mathematics to take a look at the whitepaper, compare the KB1 the source and see what can be improved on. The echo canceller code itself isn't complicated but the math is somewhat of a dark art... Just make sure to add copious plain English explanations to any changes you submit so the rest of us can keep up.

:-)