Speex dev - Dec 2005 - AEC params

Hi:

By 'no success', I mean result gets worse in terms
that, there comes in a shrill noise with high gain and
also *no* attenuation of the echo.

I am trying to validate the DC offset part. But can
you please feed some light on how does the DC offset
matter for AEC. As I have already mentioned, I tried
adding silence (from 2 ms to 10 ms) in the start of
the reference signal. This is for the condition you
have mentioned about the relationship  between
input_frame and echo_frame in the manual (Pag 19, para
3). In that case will the DC offset matter. Thanks.

Su


--------------------------------------------------------

Hi:

I have been trying to use speex for AEC.

I have recorded two voice samples separately using
audacity using the same soundcard. I reduced the gain
of one of the samples and mix this signal with the
other voice sample (saved as a raw sample, with frame
size 64 and sampling rate of 8 KHz. I use the mixed
one as a ref signal and the other unaltered signal as
the echo.

Am trying to run the testecho (with preprocessing
blocked), but no success in removing echo from the
reference.

I also tried adding small delays (5 ms 10ms), in the
reference signal, but still no success.

I tried filter lengths from 100 to 300ms but the same.

Can anyone help me sort the issue of why am not
getting any success with the echo cancellation.
Thanks.

Su

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com

> I am trying to validate the DC offset part. But can
> you please feed some light on how does the DC offset
> matter for AEC. 
I just does matter because the AEC tries to cancel it when it shouldn't
(because it fails).

	Jean-Marc

Hi:

There is no DC offset in the samples, which I verified
by opening the raw files in a editor.

Su


--- Jean-Marc Valin <Jean-Marc.Valin@USherbrooke.ca>
wrote:
> > I am trying to validate the DC offset part. But
> can
> > you please feed some light on how does the DC
> offset
> > matter for AEC. 
> 
> I just does matter because the AEC tries to cancel
> it when it shouldn't
> (because it fails).
> 
> 	Jean-Marc
> 

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com

Hi:

Can anyone give some insight on:

1) As per the Speex manual, there is a mention of the
relationship between input frame and echo frame (Pg
19, Para 2). I was wondering of what would be the
ideal delay between these two frames, so that the echo
is already fed to the canceller and the echo
cancellation filter is not inefficient.

2) Corresponding to the above mentioned delay, what
filter length should we define while creating the echo
canceller state so that we can expect the AEC to
remove or reduce the echo.

Thanks
Su

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com