Reply by October 1, 20022002-10-01
Hi Mukul,

Noise reduction before AEC appears to me as the worst solution since you
lose any hope of having a linear relation between the far end signal and
the echo enhanced by your spectral substraction.
LMS and RLS rely on that crucial linear relation.

In case of LMS, you'd better use a variable adaptation step size to avoid
divergence in presence of noise.
As far as I know, integrated solutions only deal with reducing the
residual echo thanks to noise reduction techniques.

regards,
David Mukul Bhatnagar <>
01/10/02 01:48
Please respond to echocancel To:
cc:
Subject: [echocancel] noise suppression + echo cancellation

Hi All
I had another question,
in order to make an echo cancellers performance more
robust in noise, if we integrate it with a noise
suppression algorithm- for instance Spectral
Subtraction.
What is more advisable "theoritically"
1) To put the Noise suppression algorithm before the
AEC so that the near end input to AEC is more cleaner-
reducing the possibility of divergence.
2) Or put it after the AEC, whereby , since the noise
suppression filter will probably have time varying
characteristics, would probably have to be accounted
for in the echo path for the echo canceller?

TIA
Mukul
PS>I am aware of some proposals that provide
integrated solutions, but wanted to know , how the
problem could be solved treating two individual
solutions as two separate blocks!

_____________________________________
Note: If you do a simple "reply" with your email client, only the author
of this message will receive your answer. You need to do a "reply all" if
you want your answer to be distributed to the entire group.

_____________________________________
About this discussion group:

To Join:

To Post:

To Leave:

Archives: http://www.yahoogroups.com/group/echocancel

Other DSP-Related Groups: http://www.dsprelated.com ">http://docs.yahoo.com/info/terms/



Reply by Shaw, David G David October 1, 20022002-10-01
Mukul wrote:
What is more advisable "theoritically"
1) To put the Noise suppression algorithm before the
AEC so that the near end input to AEC is more cleaner-
reducing the possibility of divergence.

(DGS) I believe you would do irreparable damage to the
signal, since spectral subtraction is clearly a very
non-linear process. I think your ability to converge
a canceller would be compromised substantially.

2) Or put it after the AEC, whereby , since the noise
suppression filter will probably have time varying
characteristics, would probably have to be accounted
for in the echo path for the echo canceller?

(DGS) Use the noise suppression as a clean up of all the
residual that you can't get rid of with cancellation. You
will have to do some clean up after the acoustic canceller
anyway.

TIA
Mukul

(DGS) Regards
Dave Shaw


Reply by Mukul Bhatnagar October 1, 20022002-10-01
Hi All
I had another question,
in order to make an echo cancellers performance more
robust in noise, if we integrate it with a noise
suppression algorithm- for instance Spectral
Subtraction.
What is more advisable "theoritically"
1) To put the Noise suppression algorithm before the
AEC so that the near end input to AEC is more cleaner-
reducing the possibility of divergence.
2) Or put it after the AEC, whereby , since the noise
suppression filter will probably have time varying
characteristics, would probably have to be accounted
for in the echo path for the echo canceller?

TIA
Mukul
PS>I am aware of some proposals that provide
integrated solutions, but wanted to know , how the
problem could be solved treating two individual
solutions as two separate blocks!