You're right - I was thinking of just doing single-delay echo cancellation. I was under the impression that in that case, processing each mic input separately makes it easier/simpler to deal with algorithmically. If what you're considering can cancel any arbitrary reflection, though, then that sounds very promising. I'll defer to you on the details since my audio knowledge does not go quite that low-level :)<br>
<br>--Andrew<br><br><div class="gmail_quote">2009/6/3 leon zadorin <span dir="ltr"><<a href="mailto:leonleon77@gmail.com">leonleon77@gmail.com</a>></span><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div class="im">On 6/4/09, leon zadorin <<a href="mailto:leonleon77@gmail.com">leonleon77@gmail.com</a>> wrote:<br>
<br>
> Laterally-speaking and talking about a given room: tell me what is the<br>
> difference between adding another mic to a mono session vs adding<br>
> another very lively reflection surface in a room? Both will cause<br>
> comb-filtering, both will add delays, both will add a spectrally<br>
> modified signal with a non-linear phase characteristics. I mean one<br>
> may abuse the above question by stating that mic's gain could be<br>
> turned way way way up (i.e. reflection surface equivalent is to<br>
> reflect w/o any energy loss) but I think even so -- the parts of<br>
> causing more reflections/delay/etc. will hold. The point being -- be<br>
> it a reflection surface or a mic -- all those things do is to filter<br>
> and delay (in an LTI fashion). In fact a reflection surface would<br>
> cause more issues because it would introduce a 'feedback' of bouncing<br>
> waves between other surfaces before the whole thing even gets to a<br>
> mic.<br>
<br>
</div>One minor addition w.r.t. the above example -- even if one was to add<br>
a mic and then speak into it really really closely :-) this would not<br>
cause any difference to the system -- because what we are mainly<br>
interested in is the<br>
"communicated data->loudspeaker->ambience->mic" path (so that<br>
external sources, other than from communicated data are not a part of<br>
the 'concern equation' -- i.e. they don't need to be canceled, in fact<br>
-- doing so would be very very very bad indeed :-) :-) :-)<br>
<br>
OK -- that's it, no more DSP talk, period :-)<br>
<br>
Kind regards<br>
<font color="#888888">Leon Zadorin.<br>
</font></blockquote></div><br>