Steve Kann wrote:
>
>> Maybe I can propose a simpler plan. Why not
implement a complete AAGC in
>> iaclient, without touching speex for now. THe
implementation can be
>> quite simple, like wscanning the waveform for the
peak value, compare it
>> to a target, and then raise or lower the mixer with
some fixed step,
>> like 0.1. At the net packet, the check will be
re-done and if the
>> previous step was not enough the mixer will be
further moved. A maximum
>> gain change step can be added to prevent rough
mixer cahnging. This
>> method is quite simple - it doesnt use signal power
estimation etc, but
>> it actually works quite well, I use similar
approach already. It is alos
>> somewhat inefficient, because it repeats the
waveform scanning that is
>> already done in agc, but this has minimal overhead,
especially in
>> narrowband. At least this can be used for a
temporary solution.
>>
>
> That is almost iaxclient does now, and what we're
trying to replace
> with something better. The difference is that the
current iaxclient
> mechanism uses speex' energy calculation, and VAD
decision. If you
> don't use a VAD decision to only change the levels when
speech is
> detected, then you end up raising the level of
background noise.
>
> -SteveK
>
Ups, you are right. My design is actually different so this
is not a
problem (I use speex NR before a combined agc/aagc).
Alex
------------------------------------------------------------
-------------
This SF.net email is sponsored by DB2 Express
Download DB2 Express C - the FREE version of DB2 express and
take
control of your XML. No limits. Just data. Click to get it
now.
http://sourcefor
ge.net/powerbar/db2/
_______________________________________________
Iaxclient-devel mailing list
Iaxclient-devel lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/iaxclie
nt-devel
|
