> 1.After the error values are computated in teh daptation process a check
-if
> filter is completely mis-adapted - is performed. How could you explain the
> criteria used for this check.
Sure, I check the correlation between what I remove and what's left
after. If that correlation is far in the negative, it means I'm actually
adding more echo instead of removing it.
> 2.what exactly is meant by over-cancellation and under-cancellation ?
This is a bit similar to point 1. I check whether I'd get better results
be multiplying what I remove by a certain factor. If I'm removing too
much (hence adding some echo with opposite phase), then it's
over-cancellation, if I'm not removing enough, then it's
under-cancellation.
> 3.Also it was seen that without the speex_preprocess call the output
quality
> is very much differnt. Is this expected ? if yes will this not increase the
> complexity of operation considerably ?
Yes, the preprocessor is able to remove a lot of the residual echo and
also performs some noise reduction.
> Is there any document which answers these querries ?
Now there's the archive :-)
Jean-Marc
--
Jean-Marc Valin <Jean-Marc.Valin@USherbrooke.ca>
Universit? de Sherbrooke
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
Url :
http://lists.xiph.org/pipermail/speex-dev/attachments/20050906/808bd1db/attachment.pgp