28 Feb
2006
28 Feb
'06
3:26 p.m.
On Tue, Feb 28, 2006 at 09:26:49AM -0600, Phil Mendelsohn wrote:
For a speech signal, such an approach could work, but it is indeed
'black magic' DSP and AFAIK there is no single app or plugin that
will do this.
The trick is to compute the 'cepstrum' of the signal - this is
the inverse FFT of the logarithm of the FFT. Then in the cepstrum
it should be possible to separate the components originating from
the original signal, and those added by the room. Remove the latter,
then do the inverse transformations. Sounds easy until you look
at the details you need to get right...
--
FA
I have a recording from a lecture. It was
recorded in a large
auditorium and of course it show in the recording. I'd like to process
the recording to get a more intimate sound. It there any way of
reducing that ballroom echo a bit? If it can be done with a ladspa
plugin so much the better.
Hi, Alex,
You can try various filtering things, but all are subject to a certain
amount of failure.
I don't know about plugin modules, but I can tell you in general, the way
you'd fix this is using "Deconvolution." The idea is simple -- it's
basically "unfiltering," but there are a couple of problems.
1: You have to have some idea what the room does to the signal.
(Possibly some info can be taken from your recording.)
2: It's really sensitive to the numbers you get from the first step -- so
even in theory (i.e., on paper) it doesn't work all that well. In
practice it's pretty hard to make work -- this is really "black magic"
DSP. [though there's no such thing.]