Speex dev - Jul 2011 - Extracting Coefficients from the AEC

Steve U.,

Thanks for the quick response. I will look into OSLEC for the Hybrid case. With
regards to using the
last known good set of coefficients for re-use, I am assuming that any changes
in the echo path would
be minimal to none (in the hybrid case of course) and therefore if my system was
reset, only a small
amount of convergence would be needed if at all. Of course I am basing my
assumption on my small amount
of knowledge on adaptive signal processing theory and am assuming that the
filter is modeling the FIR
response of the path and isn't necessarily dependant on the signal. This
will be my first
implementation of an echo canceller, so I guess I will soo find out:)

Thanks,

Doug C.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 5212 bytes
Desc: not available
Url :
http://lists.xiph.org/pipermail/speex-dev/attachments/20110727/2bbf8c0e/attachment.bin

On 07/28/2011 05:21 AM, Cooper, Douglas A CIV NAVAIR, 4.6.2.3
wrote:
> Steve U.,
>
> Thanks for the quick response. I will look into OSLEC for the Hybrid case.
With regards to using the
> last known good set of coefficients for re-use, I am assuming that any
changes in the echo path would
> be minimal to none (in the hybrid case of course) and therefore if my
system was reset, only a small
> amount of convergence would be needed if at all. Of course I am basing my
assumption on my small amount
> of knowledge on adaptive signal processing theory and am assuming that the
filter is modeling the FIR
> response of the path and isn't necessarily dependant on the signal.
This will be my first
> implementation of an echo canceller, so I guess I will soo find out:)
Only your local end hybrids (in your terminal and the local exchange) 
are reasonably constant. The far end phone and egress exchange hybrid 
are different each time. Are you assuming they are insignificant? If you 
are, its a mistake. If one call requires a huge allowance for a far end 
echo, and the next doesn't, starting the second call with the coeffs 
from the first can be worse than starting from scratch.

Some people fool themselves by making many calls, but always to the same 
far end phone. You need some diversity. Beware of testing with only long 
distance calls, too. These usually have near zero far end echo, because 
there are usually in-network echo cancellers removing the echo before 
you ever see it. Local calls are not cancelled by the network.

Steve