> But an FIR or IIR filter can't compare coherent signal to
> incoherent noise and enhance multiple signals while
> suppressing the noise between them.
The textbook I referred to is available from Amazon. You can take it up
with the author if you don't agree with the mathematical derivation or
operation of a Wiener filter. It is written in English, which is my native
tongue :-)
I bought it (at the good suggestion of someone, can't recall who, on the
list), the last time we had this getting-very-old discussion of what might
or might not make good de-noiseing algorithms. The book is an excellent
reference. And very informative reading, particularly since I have a
business interest in continuous speech recognition in the presence of (a lot
of) noise.
I am NOT an expert in noise reduction but I can read, and generally do not
misinterpret clear statements. " .. For additive noise, the Wiener filter
attenuates each frequency component in proportion to an estimate of the
signal to noise ratio. ... For additive white noise, the Wiener filter
response broadly follows the signal spectrum." Further, ".. [for noise and
signal spectra which overlap] it is not possible to completely separate the
signal from the noise. However, the effects of the noise can be reduced by
using a Wiener filer that ATTENUATES EACY NOISY SIGNAL FREQUENCY [emphasis
added] in proportion to an estimate of the signal-to-noise ratio."
I'd be happy to scan and send you Figure 6.5, page 179 of the referenced
text which describes, graphically, precisely what I said. I'd also be
happy to send the block diagram of the Wiener filter, showing how the error
component is used to adapt the coeffiecients of an FIR filter, which is the
basis for the discussion on Wiener filters in the referenced text.
At some point, perhaps we should all just go out and beat ourselves with a
stick :-)
Grant/NQ5T
_______________________________________________
TenTec mailing list
TenTec@contesting.com
http://lists.contesting.com/mailman/listinfo/tentec
|