Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
Iterative Match Filterings
29 views
Skip to first unread message
Bret Cahill
Apr 3, 2012, 1:25:33 AM4/3/12
to
Instead of just taking the convolution of a signal with the reference
or kernel once and then taking the square root in the frequency
domain, why not match filter with the reference again and again to
recover the original clean signal?
Well, you _can_ do that if a wave form is all you want, but you can
get that from the ref. alone _anyway_.
If you want to keep the original magnitude of the signal, however, and
if the kernel is a different magnitude than the original signal, then
you will only be working your way back to the kernel with each
iteration.
The solution here is to convolve the signal with the kernel and then
the kernel with the kernel and then comparing the magnitudes of those
two convolutions.
Apply that factor to correct the magnitude of the filtered signal.
Bret Cahill
or kernel once and then taking the square root in the frequency
domain, why not match filter with the reference again and again to
recover the original clean signal?
Well, you _can_ do that if a wave form is all you want, but you can
get that from the ref. alone _anyway_.
If you want to keep the original magnitude of the signal, however, and
if the kernel is a different magnitude than the original signal, then
you will only be working your way back to the kernel with each
iteration.
The solution here is to convolve the signal with the kernel and then
the kernel with the kernel and then comparing the magnitudes of those
two convolutions.
Apply that factor to correct the magnitude of the filtered signal.
Bret Cahill
Bret Cahill
Apr 3, 2012, 10:29:16 AM4/3/12
to
Is there anything to be gained with iterative match filterings as far
as recovering the correct magnitude is concerned?
Bret Cahill
Les Cargill
Apr 3, 2012, 1:52:12 PM4/3/12
to
of any filter is almost guaranteed to change the magnitude of the result.
I've played with normalization* of the output signal back to the
magnitude of the original and I don't detect any general pattern.
*almost always audio signals, and the measure of magnitude is
almost always the RMS of the vector in question...
If the mag. of the result is 0.5 dB down, then add 0.5 gain... this
may or may not mean you can change the filter to do that
for you... for convolution kernels, I belive it *can*, but IIR
filters almost certainly don't work that way...
--
Les Cargill
Tim Wescott
Apr 3, 2012, 2:59:54 PM4/3/12
to
filtering degrades the output.
If doing matched filtering _isn't_ optimal in the sense that you desire,
then you need to choose a different filter, designed to an optimality
constraint that matches your desire.
Trying to solve a problem by randomly picking terminology out of a hat,
then throwing the result at the problem, is not optimal.
--
My liberal friends think I'm a conservative kook.
My conservative friends think I'm a liberal kook.
Why am I not happy that they have found common ground?
Tim Wescott, Communications, Control, Circuits & Software
http://www.wescottdesign.com
Fred Marshall
Apr 3, 2012, 4:07:53 PM4/3/12
to
On 4/2/2012 10:25 PM, Bret Cahill wrote:
> Instead of just taking the convolution of a signal with the reference
> or kernel once and then taking the square root in the frequency
> domain, why not match filter with the reference again and again to
> recover the original clean signal?
Perhaps you might consider:
> Instead of just taking the convolution of a signal with the reference
> or kernel once and then taking the square root in the frequency
> domain, why not match filter with the reference again and again to
> recover the original clean signal?
"What is the purpose of applying a matched filter?"
I think you will find that the purpose is to enhance the SNR of the
*detection of the presence* of the original signal in noise.
Just consider that the calculation is that of an inner product between
two vectors. The inner product is a scalar.
There is a scalar output in time for each sample in time.
So, the objective is to align the vectors in time such that it
statistically generates a maximum output when the alignment is best.
This is not about extracting a waveform. It's about extracting a maximum.
I think it's fair to say that by extracting a maximum of reasonable
value, the underlying clean waveform in noise is *implied*. But that's
not "extraction" exactly is it? That is to say: if you apply some
filter and want the original waveform to emerge then that's different
than saying you want to detect the presence of the waveform in noise.
There is a time/bandwidth tradeoff here as well.
So, a very narrowband filter could be a matched filter for a suitably
long sinusoid but with very poor temporal resolution as to "when did it
happen?" But what about a short sinusoidal pulse? The temporal
resolution is improved at the expense of frequency resolution. So, if
we ask: "Did a sinusoidal pulse arrive at a particular time with a
particular frequency?" there is ambiguity regarding both time and
frequency when a maximum is detected.
By extension, one might imply or even *measure* the apparent amplitude
of the apparently present waveform in some context involving
calibration, etc. But this gets fuzzed up if there are frequency
variations and is dealt with by using multiple matched filters centered
at different frequencies. .... and more beyond that.
Fred
Bret Cahill
Apr 3, 2012, 11:30:59 PM4/3/12
to
> >> Instead of just taking the convolution of a signal with the reference
> >> or kernel once and then taking the square root in the frequency
> >> domain, why not match filter with the reference again and again to
> >> recover the original clean signal?
>
> >> Well, you _can_ do that if a wave form is all you want, but you can
> >> get that from the ref. alone _anyway_.
>
> >> If you want to keep the original magnitude of the signal, however, and
> >> if the kernel is a different magnitude than the original signal, then
> >> you will only be working your way back to the kernel with each
> >> iteration.
>
> >> The solution here is to convolve the signal with the kernel and then
> >> the kernel with the kernel and then comparing the magnitudes of those
> >> two convolutions.
>
> >> Apply that factor to correct the magnitude of the filtered signal.
>
> > One question is:
>
> > Is there anything to be gained with iterative match filterings as far
> > as recovering the correct magnitude is concerned?
>
> > Bret Cahill
>
> This answer is loaded with observer bias, but I find that application
> of any filter is almost guaranteed to change the magnitude of the result.
That shouldn't be a problem here as long as the % change is the same
> >> or kernel once and then taking the square root in the frequency
> >> domain, why not match filter with the reference again and again to
> >> recover the original clean signal?
>
> >> Well, you _can_ do that if a wave form is all you want, but you can
> >> get that from the ref. alone _anyway_.
>
> >> If you want to keep the original magnitude of the signal, however, and
> >> if the kernel is a different magnitude than the original signal, then
> >> you will only be working your way back to the kernel with each
> >> iteration.
>
> >> The solution here is to convolve the signal with the kernel and then
> >> the kernel with the kernel and then comparing the magnitudes of those
> >> two convolutions.
>
> >> Apply that factor to correct the magnitude of the filtered signal.
>
> > One question is:
>
> > Is there anything to be gained with iterative match filterings as far
> > as recovering the correct magnitude is concerned?
>
> > Bret Cahill
>
> This answer is loaded with observer bias, but I find that application
> of any filter is almost guaranteed to change the magnitude of the result.
for both filtered signals. The reason for this is because the result
is always the quotient of two filtered signals.
This works out ok for just the convolution but if you complete the
match filtering process by taking de convolutions to recover the
original wave forms, then the magnitudes of both signals move toward
the magnitude of the kernel and the quotient moves closer to 1.
> I've played with normalization* of the output signal back to the
> magnitude of the original and I don't detect any general pattern.
convolutions.
Trying to recover the magnitudes of the original signals from the
match filtered signals requires 3 additional steps:
1. match filtering the kernel with itself and then,
2. comparing that with each match filtered signal for a factor.
3. applying that correction factor to it's respective signal.
Bret Cahill
Apr 3, 2012, 11:42:16 PM4/3/12
to
> > Instead of just taking the convolution of a signal with the reference
> > or kernel once and then taking the square root in the frequency
> > domain, why not match filter with the reference again and again to
> > recover the original clean signal?
>
> Perhaps you might consider:
> "What is the purpose of applying a matched filter?"
To get the magnitudes of the clean signals.
> > or kernel once and then taking the square root in the frequency
> > domain, why not match filter with the reference again and again to
> > recover the original clean signal?
>
> Perhaps you might consider:
> "What is the purpose of applying a matched filter?"
Taking the convolution of each signal with the reference low pass
filters and eliminates multiple crossings problems.
If a quotient of two signals convolved with the same kernel is taken
then any magnitude change is the same in both numerator and
denominator so it cancels in the quotient.
If you deconvolve the convolution to recover the original wave form --
basically completing the match filtering process -- then taking the
quotient will not cancel the magnitude error that was introduced by
the convolution.
Tim Wescott
Apr 4, 2012, 5:44:09 PM4/4/12
to
On Tue, 03 Apr 2012 20:42:16 -0700, Bret Cahill wrote:
>> > Instead of just taking the convolution of a signal with the reference
>> > or kernel once and then taking the square root in the frequency
>> > domain, why not match filter with the reference again and again to
>> > recover the original clean signal?
>>
>> Perhaps you might consider:
>> "What is the purpose of applying a matched filter?"
>
> To get the magnitudes of the clean signals.
Yes, we already established that _you_ want the filter to do that. What
>> > Instead of just taking the convolution of a signal with the reference
>> > or kernel once and then taking the square root in the frequency
>> > domain, why not match filter with the reference again and again to
>> > recover the original clean signal?
>>
>> Perhaps you might consider:
>> "What is the purpose of applying a matched filter?"
>
> To get the magnitudes of the clean signals.
_Fred_ is saying (and, for that matter, I also) is
>>> What does the MATH say that a matched filter is designed for <<<
Because if you ain't applying a filter to the purpose for which it is
designed, and hoping that a high concentration of buzz-words will make
things operate well, then you ain't doing engineering -- you're doing
magic, and the next step is to go buy some live chickens to sacrifice.
Bret Cahill
Apr 4, 2012, 7:24:37 PM4/4/12
to
> Because if you ain't applying a filter to the purpose for which it is
> designed
If you can't figure out how to use convolutions and match filtering to
> designed
determine magnitudes of noisy signals you need to start a thread
entitled "A Scholarly Enumeration of Filters & Their Purposes."
Maybe you can get it in Wikipedia!
Bret Cahill
Tim Wescott
Apr 4, 2012, 7:46:54 PM4/4/12
to
On Wed, 04 Apr 2012 16:24:37 -0700, Bret Cahill wrote:
>> Because if you ain't applying a filter to the purpose for which it is
>> designed
>
> If you can't figure out how to use convolutions and match filtering to
> determine magnitudes of noisy signals you need to start a thread
> entitled "A Scholarly Enumeration of Filters & Their Purposes."
(A): _I_ can.
>> Because if you ain't applying a filter to the purpose for which it is
>> designed
>
> If you can't figure out how to use convolutions and match filtering to
> determine magnitudes of noisy signals you need to start a thread
> entitled "A Scholarly Enumeration of Filters & Their Purposes."
(B): You don't even have a clue what you're saying.
(C): If you can't ask a misguided question and then accept
help getting your _question_ on track, then you shouldn't
ask questions.
Bret Cahill
Apr 4, 2012, 11:45:00 PM4/4/12
to
> >> Because if you ain't applying a filter to the purpose for which it is
> >> designed
>
> > If you can't figure out how to use convolutions and match filtering to
> > determine magnitudes of noisy signals you need to start a thread
> > entitled "A Scholarly Enumeration of Filters & Their Purposes."
>
> (A): _I_ can.
You'll need to _demonstrate_ that with your post "A Scholarly
> >> designed
>
> > If you can't figure out how to use convolutions and match filtering to
> > determine magnitudes of noisy signals you need to start a thread
> > entitled "A Scholarly Enumeration of Filters & Their Purposes."
>
> (A): _I_ can.
Enumeration of Filters & Their Purposes."
Bret Cahill
Bret Cahill
Apr 5, 2012, 6:33:20 AM4/5/12
to
signal and divide that by the magnitude of the kernel match filtered
with itself.
Bret Cahill
Martin Brown
Apr 5, 2012, 6:59:49 AM4/5/12
to
On 05/04/2012 00:24, Bret Cahill wrote:
>> Because if you ain't applying a filter to the purpose for which it is
>> designed
>
> If you can't figure out how to use convolutions and match filtering to
> determine magnitudes of noisy signals you need to start a thread
> entitled "A Scholarly Enumeration of Filters& Their Purposes."
>> Because if you ain't applying a filter to the purpose for which it is
>> designed
>
> If you can't figure out how to use convolutions and match filtering to
> determine magnitudes of noisy signals you need to start a thread
>
> Maybe you can get it in Wikipedia!
>
> Bret Cahill
Wiki is a bit weak on iterative deconvolution algorithms.
> Maybe you can get it in Wikipedia!
>
> Bret Cahill
The sort of thing you are asking about, or rather a version that
actually works in practice was invented in 1974 by Hogbom as the CLEAN
algorithm for radio astronomy. His original paper is online at:
http://www.astro.rug.nl/~vdhulst/SignalProcessing/Project1_data/clean_hogbom.pdf
Variants of this algorithm are still used today in aperture synthesis
although I prefer other deconvolution algorithms myself.
--
Regards,
Martin Brown
Bret Cahill
Apr 5, 2012, 10:37:12 AM4/5/12
to
> >> Because if you ain't applying a filter to the purpose for which it is
> >> designed
>
> > If you can't figure out how to use convolutions and match filtering to
> > determine magnitudes of noisy signals you need to start a thread
> > entitled "A Scholarly Enumeration of Filters& Their Purposes."
>
> > Maybe you can get it in Wikipedia!
>
> > Bret Cahill
>
> Wiki is a bit weak on iterative deconvolution algorithms.
>
> The sort of thing you are asking about, or rather a version that
> actually works in practice was invented in 1974 by Hogbom as the CLEAN
> algorithm for radio astronomy. His original paper is online at:
>
> http://www.astro.rug.nl/~vdhulst/SignalProcessing/Project1_data/clean...
> >> designed
>
> > If you can't figure out how to use convolutions and match filtering to
> > determine magnitudes of noisy signals you need to start a thread
> > entitled "A Scholarly Enumeration of Filters& Their Purposes."
>
> > Maybe you can get it in Wikipedia!
>
> > Bret Cahill
>
> Wiki is a bit weak on iterative deconvolution algorithms.
>
> The sort of thing you are asking about, or rather a version that
> actually works in practice was invented in 1974 by Hogbom as the CLEAN
> algorithm for radio astronomy. His original paper is online at:
>
>
> Variants of this algorithm are still used today in aperture synthesis
> although I prefer other deconvolution algorithms myself.
In match filtering the deconvolution step to recover the original
> Variants of this algorithm are still used today in aperture synthesis
> although I prefer other deconvolution algorithms myself.
[filtered]wave form isn't complicated. Just take the square root in
the frequency domain, IMSQRT in Excel.
Of course, if the magnitude of the kernel is different than the
magnitude of the signal -- the general case -- then the magnitude of
the recovered signal will be the sqrt of the product of the 2
magnitudes.
If you want to get the original magnitude of the signal just square
the mag. of the filtered signal and then divide by the mag. of the
kernel, or, to keep errors equal, divide by the kernel match filtered
with itself.
This was done 7 decades ago, wasn't it?
Bret Cahill
Martin Brown
Apr 5, 2012, 11:49:29 AM4/5/12
to
On 05/04/2012 15:37, Bret Cahill wrote:
>>>> Because if you ain't applying a filter to the purpose for which it is
>>>> designed
>>
>>> If you can't figure out how to use convolutions and match filtering to
>>> determine magnitudes of noisy signals you need to start a thread
>>> entitled "A Scholarly Enumeration of Filters& Their Purposes."
>>
>>> Maybe you can get it in Wikipedia!
>>
>>> Bret Cahill
>>
>> Wiki is a bit weak on iterative deconvolution algorithms.
>>
>> The sort of thing you are asking about, or rather a version that
>> actually works in practice was invented in 1974 by Hogbom as the CLEAN
>> algorithm for radio astronomy. His original paper is online at:
>>
>> http://www.astro.rug.nl/~vdhulst/SignalProcessing/Project1_data/clean...
>>
>> Variants of this algorithm are still used today in aperture synthesis
>> although I prefer other deconvolution algorithms myself.
>
> In match filtering the deconvolution step to recover the original
> [filtered]wave form isn't complicated. Just take the square root in
> the frequency domain, IMSQRT in Excel.
Anyone who says deconvolution isn't complicated demonstrably does not
>>>> Because if you ain't applying a filter to the purpose for which it is
>>>> designed
>>
>>> If you can't figure out how to use convolutions and match filtering to
>>> determine magnitudes of noisy signals you need to start a thread
>>> entitled "A Scholarly Enumeration of Filters& Their Purposes."
>>
>>> Maybe you can get it in Wikipedia!
>>
>>> Bret Cahill
>>
>> Wiki is a bit weak on iterative deconvolution algorithms.
>>
>> The sort of thing you are asking about, or rather a version that
>> actually works in practice was invented in 1974 by Hogbom as the CLEAN
>> algorithm for radio astronomy. His original paper is online at:
>>
>> http://www.astro.rug.nl/~vdhulst/SignalProcessing/Project1_data/clean...
>>
>> Variants of this algorithm are still used today in aperture synthesis
>> although I prefer other deconvolution algorithms myself.
>
> In match filtering the deconvolution step to recover the original
> [filtered]wave form isn't complicated. Just take the square root in
> the frequency domain, IMSQRT in Excel.
have the first clue about the subject of signal processing. Convolution
is easy but deconvolution is usually a very difficult inverse problem.
> Of course, if the magnitude of the kernel is different than the
> magnitude of the signal -- the general case -- then the magnitude of
> the recovered signal will be the sqrt of the product of the 2
> magnitudes.
>
> If you want to get the original magnitude of the signal just square
> the mag. of the filtered signal and then divide by the mag. of the
> kernel, or, to keep errors equal, divide by the kernel match filtered
> with itself.
>
> This was done 7 decades ago, wasn't it?
determine what "this" you are referring to. If you mean noise adaptive
Kalman filtering then it was actually first discovered in the 1880's by
Thiele but published in Danish so no-one noticed it at all. His bad luck
- he also explained Brownian motion well before Einstein.
He came up recently in one of the periodic netkooks antisemitic rants
against Einstein which flare up in sci.physics from time to time.
http://groups.google.com/group/sci.physics/msg/586291ff896f9fd8?hl=en
Only belatedly is poor Thiele's earlier work getting any recognition.
http://en.wikipedia.org/wiki/Thorvald_Thiele
Kalman filters are a heck of a lot of work to implement using pen and
pencil in the days before digital computers. He was too far ahead of his
time to get any recognition for his breakthrough in this field.
--
Regards,
Martin Brown
Bret Cahill
Apr 5, 2012, 12:27:23 PM4/5/12
to
> >>>> Because if you ain't applying a filter to the purpose for which it is
> >>>> designed
>
> >>> If you can't figure out how to use convolutions and match filtering to
> >>> determine magnitudes of noisy signals you need to start a thread
> >>> entitled "A Scholarly Enumeration of Filters& Their Purposes."
>
> >>> Maybe you can get it in Wikipedia!
>
> >>> Bret Cahill
>
> >> Wiki is a bit weak on iterative deconvolution algorithms.
>
> >> The sort of thing you are asking about, or rather a version that
> >> actually works in practice was invented in 1974 by Hogbom as the CLEAN
> >> algorithm for radio astronomy. His original paper is online at:
>
> >>http://www.astro.rug.nl/~vdhulst/SignalProcessing/Project1_data/clean...
>
> >> Variants of this algorithm are still used today in aperture synthesis
> >> although I prefer other deconvolution algorithms myself.
>
> > In match filtering the deconvolution step to recover the original
> > [filtered]wave form isn't complicated. Just take the square root in
> > the frequency domain, IMSQRT in Excel.
>
> Anyone who says deconvolution isn't complicated demonstrably does not
> have the first clue about the subject of signal processing.
The issue here is the special kind of convolution used in match
> >>>> designed
>
> >>> If you can't figure out how to use convolutions and match filtering to
> >>> determine magnitudes of noisy signals you need to start a thread
> >>> entitled "A Scholarly Enumeration of Filters& Their Purposes."
>
> >>> Maybe you can get it in Wikipedia!
>
> >>> Bret Cahill
>
> >> Wiki is a bit weak on iterative deconvolution algorithms.
>
> >> The sort of thing you are asking about, or rather a version that
> >> actually works in practice was invented in 1974 by Hogbom as the CLEAN
> >> algorithm for radio astronomy. His original paper is online at:
>
> >>http://www.astro.rug.nl/~vdhulst/SignalProcessing/Project1_data/clean...
>
> >> Variants of this algorithm are still used today in aperture synthesis
> >> although I prefer other deconvolution algorithms myself.
>
> > In match filtering the deconvolution step to recover the original
> > [filtered]wave form isn't complicated. Just take the square root in
> > the frequency domain, IMSQRT in Excel.
>
> Anyone who says deconvolution isn't complicated demonstrably does not
> have the first clue about the subject of signal processing.
filtering.
If you want to change the issue to convolutions generally feel free to
start another thread.
> Convolution
> is easy but deconvolution is usually
Does this include the convolution of a function with itself?
Remember, no dodging.
> a very difficult inverse problem.
itself.
In that case the deconvolution just requires taking the square root in
the frequency domain.
> > Of course, if the magnitude of the kernel is different than the
> > magnitude of the signal -- the general case -- then the magnitude of
> > the recovered signal will be the sqrt of the product of the 2
> > magnitudes.
> > If you want to get the original magnitude of the signal just square
> > the mag. of the filtered signal and then divide by the mag. of the
> > kernel, or, to keep errors equal, divide by the kernel match filtered
> > with itself.
>
> > This was done 7 decades ago, wasn't it?
>
> Your description is so vague and woolly that
be able to show it on Excel.
WARNING: THIS MAY ALREADY BE ON A LINK SOMEWHERE! STOP DIGGING AND
START GOOGLING!
Bret Cahill
Eric Jacobsen
Apr 5, 2012, 2:30:36 PM4/5/12
to
On Thu, 5 Apr 2012 09:27:23 -0700 (PDT), Bret Cahill
<BretC...@peoplepc.com> wrote:
>> >>>> Because if you ain't applying a filter to the purpose for which it i=
>s
>> >>>> designed
>>
>> >>> If you can't figure out how to use convolutions and match filtering t=
different? If you mean matched filtering then no special kind of
convolution is needed. You need to clarify what you mean. You're
getting flak back because the way you're describing things doesn't
make much sense.
Eric Jacobsen
Anchor Hill Communications
www.anchorhill.com
<BretC...@peoplepc.com> wrote:
>> >>>> Because if you ain't applying a filter to the purpose for which it i=
>s
>> >>>> designed
>>
>> >>> If you can't figure out how to use convolutions and match filtering t=
>o
>> >>> determine magnitudes of noisy signals you need to start a thread
>> >>> entitled "A Scholarly Enumeration of Filters& =A0 =A0Their Purposes."
>> >>> determine magnitudes of noisy signals you need to start a thread
>>
>> >>> Maybe you can get it in Wikipedia!
>>
>> >>> Bret Cahill
>>
>> >> Wiki is a bit weak on iterative deconvolution algorithms.
>>
>> >> The sort of thing you are asking about, or rather a version that
>> >> actually works in practice was invented in 1974 by Hogbom as the CLEAN
>> >> algorithm for radio astronomy. His original paper is online at:
>>
>> >>http://www.astro.rug.nl/~vdhulst/SignalProcessing/Project1_data/clean..=
>> >>> Maybe you can get it in Wikipedia!
>>
>> >>> Bret Cahill
>>
>> >> Wiki is a bit weak on iterative deconvolution algorithms.
>>
>> >> The sort of thing you are asking about, or rather a version that
>> >> actually works in practice was invented in 1974 by Hogbom as the CLEAN
>> >> algorithm for radio astronomy. His original paper is online at:
>>
>.
>>
>> >> Variants of this algorithm are still used today in aperture synthesis
>> >> although I prefer other deconvolution algorithms myself.
>>
>> > In match filtering the deconvolution step to recover the original
>> > [filtered]wave form isn't complicated. =A0Just take the square root in
>>
>> >> Variants of this algorithm are still used today in aperture synthesis
>> >> although I prefer other deconvolution algorithms myself.
>>
>> > In match filtering the deconvolution step to recover the original
>> > the frequency domain, IMSQRT in Excel.
>>
>> Anyone who says deconvolution isn't complicated demonstrably does not
>> have the first clue about the subject of signal processing.
>
>The issue here is the special kind of convolution used in match
>filtering.
Do you mean matched filtering or is "match" filtering something
>>
>> Anyone who says deconvolution isn't complicated demonstrably does not
>> have the first clue about the subject of signal processing.
>
>The issue here is the special kind of convolution used in match
>filtering.
different? If you mean matched filtering then no special kind of
convolution is needed. You need to clarify what you mean. You're
getting flak back because the way you're describing things doesn't
make much sense.
Eric Jacobsen
Anchor Hill Communications
www.anchorhill.com
glen herrmannsfeldt
Apr 5, 2012, 4:55:44 PM4/5/12
to
In comp.dsp Martin Brown <|||newspam|||@nezumi.demon.co.uk> wrote:
(snip)
gives less than useful results. As usual, the non-linear case is
more complicated than the linear case.
I have previously recommended "Deconvolution of Images and Spectra"
by Jansson. Rumors are that it is now available in an affordable
paperback reprint. Otherwise it is pretty expensive.
-- glen
(snip)
> Anyone who says deconvolution isn't complicated demonstrably does not
> have the first clue about the subject of signal processing. Convolution
> is easy but deconvolution is usually a very difficult inverse problem.
Well, linear deconvolution isn't complicated, but it also often
> have the first clue about the subject of signal processing. Convolution
> is easy but deconvolution is usually a very difficult inverse problem.
gives less than useful results. As usual, the non-linear case is
more complicated than the linear case.
I have previously recommended "Deconvolution of Images and Spectra"
by Jansson. Rumors are that it is now available in an affordable
paperback reprint. Otherwise it is pretty expensive.
-- glen
Bret Cahill
Apr 5, 2012, 7:46:11 PM4/5/12
to
Bret Cahill
Eric Jacobsen
Apr 5, 2012, 8:48:13 PM4/5/12
to
special kind of convolution, only the ordinary, mundane kind. There's
no deconvolution involved.
You need to explain what you really mean, if you can. I suspect
you're confused about it.
Bret Cahill
Apr 5, 2012, 11:51:49 PM4/5/12
to
> which requires no
> special kind of convolution, only the ordinary, mundane kind. There's
> no deconvolution involved.
Bret Cahill
Eric Jacobsen
Apr 6, 2012, 1:29:05 PM4/6/12
to
On Thu, 5 Apr 2012 20:51:49 -0700 (PDT), Bret Cahill
<BretC...@peoplepc.com> wrote:
>On Apr 5, 5:48=A0pm, eric.jacob...@ieee.org (Eric Jacobsen) wrote:
>> On Thu, 5 Apr 2012 16:46:11 -0700 (PDT), Bret Cahill
>>
>>
>>
>>
>>
>> <BretCah...@peoplepc.com> wrote:
>> >> >> >>>> Because if you ain't applying a filter to the purpose for whic=
>h it i=3D
>> >> >s
>> >> >> >>>> designed
>>
>> >> >> >>> If you can't figure out how to use convolutions and match filte=
>ring t=3D
>> >> >o
>> >> >> >>> determine magnitudes of noisy signals you need to start a threa=
>d
>> >> >> >>> entitled "A Scholarly Enumeration of Filters& =3DA0 =3DA0Their =
>lean..=3D
>> >> >.
>>
>> >> >> >> Variants of this algorithm are still used today in aperture synt=
matched filtering process, so perhaps you can understand the confusion
among the readers about what you're talking about.
>> which requires no
>> special kind of convolution, only the ordinary, mundane kind. =A0There's
filters are used for all sorts of tasks, from distance measuring (in
radar, sonar, etc.) to recovering modulated data (in a communication
system) to all sorts of other things. In a communication system the
"original waveform" might be the modulated data, but that doesn't
really translate well to a radar.
Only you seem to know what you mean, and you need to provide something
better than gibberish if you want others to contribute meaningfully to
whatever questions you might have.
<BretC...@peoplepc.com> wrote:
>On Apr 5, 5:48=A0pm, eric.jacob...@ieee.org (Eric Jacobsen) wrote:
>> On Thu, 5 Apr 2012 16:46:11 -0700 (PDT), Bret Cahill
>>
>>
>>
>>
>>
>> <BretCah...@peoplepc.com> wrote:
>h it i=3D
>> >> >s
>> >> >> >>>> designed
>>
>> >> >> >>> If you can't figure out how to use convolutions and match filte=
>ring t=3D
>> >> >o
>> >> >> >>> determine magnitudes of noisy signals you need to start a threa=
>d
>> >> >> >>> entitled "A Scholarly Enumeration of Filters& =3DA0 =3DA0Their =
>Purposes."
>>
>> >> >> >>> Maybe you can get it in Wikipedia!
>>
>> >> >> >>> Bret Cahill
>>
>> >> >> >> Wiki is a bit weak on iterative deconvolution algorithms.
>>
>> >> >> >> The sort of thing you are asking about, or rather a version that
>> >> >> >> actually works in practice was invented in 1974 by Hogbom as the=
>>
>> >> >> >>> Maybe you can get it in Wikipedia!
>>
>> >> >> >>> Bret Cahill
>>
>> >> >> >> Wiki is a bit weak on iterative deconvolution algorithms.
>>
>> >> >> >> The sort of thing you are asking about, or rather a version that
> CLEAN
>> >> >> >> algorithm for radio astronomy. His original paper is online at:
>>
>> >> >> >>http://www.astro.rug.nl/~vdhulst/SignalProcessing/Project1_data/c=
>> >> >> >> algorithm for radio astronomy. His original paper is online at:
>>
>lean..=3D
>> >> >.
>>
>> >> >> >> Variants of this algorithm are still used today in aperture synt=
>hesis
>> >> >> >> although I prefer other deconvolution algorithms myself.
>>
>> >> >> > In match filtering the deconvolution step to recover the original
>> >> >> > [filtered]wave form isn't complicated. =3DA0Just take the square =
>> >> >> >> although I prefer other deconvolution algorithms myself.
>>
>> >> >> > In match filtering the deconvolution step to recover the original
>root in
>> >> >> > the frequency domain, IMSQRT in Excel.
>>
>> >> >> Anyone who says deconvolution isn't complicated demonstrably does n=
>> >> >> > the frequency domain, IMSQRT in Excel.
>>
>ot
>> >> >> have the first clue about the subject of signal processing.
>>
>> >> >The issue here is the special kind of convolution used in match
>> >> >filtering.
>>
>> >> Do you mean matched filtering or is "match" filtering something
>> >> different?
>>
>> >North filter.
>>
>> Far more commonly known as a "matched filter",
>
>How long did it take you to figure that out?
Far longer than it should have. What you've been describing isn't a
>> >> >> have the first clue about the subject of signal processing.
>>
>> >> >The issue here is the special kind of convolution used in match
>> >> >filtering.
>>
>> >> Do you mean matched filtering or is "match" filtering something
>> >> different?
>>
>> >North filter.
>>
>> Far more commonly known as a "matched filter",
>
>How long did it take you to figure that out?
matched filtering process, so perhaps you can understand the confusion
among the readers about what you're talking about.
>> which requires no
>> special kind of convolution, only the ordinary, mundane kind. =A0There's
>> no deconvolution involved.
>
>How is the original waveform recovered with a North Filter?
You need to be clear what you mean by "original waveform". Matched
>
>How is the original waveform recovered with a North Filter?
filters are used for all sorts of tasks, from distance measuring (in
radar, sonar, etc.) to recovering modulated data (in a communication
system) to all sorts of other things. In a communication system the
"original waveform" might be the modulated data, but that doesn't
really translate well to a radar.
Only you seem to know what you mean, and you need to provide something
better than gibberish if you want others to contribute meaningfully to
whatever questions you might have.
robert bristow-johnson
Apr 6, 2012, 1:43:49 PM4/6/12
to
On 4/5/12 11:51 PM, Bret Cahill wrote:
> On Apr 5, 5:48 pm, eric.jacob...@ieee.org (Eric Jacobsen) wrote:
>> On Thu, 5 Apr 2012 16:46:11 -0700 (PDT), Bret Cahill
>>
> On Apr 5, 5:48 pm, eric.jacob...@ieee.org (Eric Jacobsen) wrote:
>> On Thu, 5 Apr 2012 16:46:11 -0700 (PDT), Bret Cahill
>>
>>>> Do you mean matched filtering or is "match" filtering something
>>>> different?
>>
>>> North filter.
>>
>> Far more commonly known as a "matched filter",
>
> How long did it take you to figure that out?
>
>> which requires no
>> special kind of convolution, only the ordinary, mundane kind. There's
>> no deconvolution involved.
>
> How is the original waveform recovered with a North Filter?
Bret, can you stick with the common semantic? before today, i have
>>>> different?
>>
>>> North filter.
>>
>> Far more commonly known as a "matched filter",
>
> How long did it take you to figure that out?
>
>> which requires no
>> special kind of convolution, only the ordinary, mundane kind. There's
>> no deconvolution involved.
>
> How is the original waveform recovered with a North Filter?
never heard of a "North filter". i am assuming from Eric's response and
yours, that it is synonymous with "matched filter".
a matched filter is not about recovering an original waveform. it is
about *detecting* it (among some other "original waveforms") in the
presence of noise or other crap that might obscure it.
--
r b-j r...@audioimagination.com
"Imagination is more important than knowledge."
Bret Cahill
Apr 6, 2012, 2:03:36 PM4/6/12
to
> >>>> Do you mean matched filtering or is "match" filtering something
> >>>> different?
>
> >>> North filter.
>
> >> Far more commonly known as a "matched filter",
>
> > How long did it take you to figure that out?
>
> >> which requires no
> >> special kind of convolution, only the ordinary, mundane kind. There's
> >> no deconvolution involved.
>
> > How is the original waveform recovered with a North Filter?
>
> Bret, can you stick with the common semantic? before today, i have
> never heard of a "North filter".
Is anyone still whining about "match filter" or "matched filter", both
> >>>> different?
>
> >>> North filter.
>
> >> Far more commonly known as a "matched filter",
>
> > How long did it take you to figure that out?
>
> >> which requires no
> >> special kind of convolution, only the ordinary, mundane kind. There's
> >> no deconvolution involved.
>
> > How is the original waveform recovered with a North Filter?
>
> Bret, can you stick with the common semantic? before today, i have
> never heard of a "North filter".
of which google up tens of thousands of on point hits in less time
than it takes to play trifling word games.
> i am assuming from Eric's response and
> yours, that it is synonymous with "matched filter".
> a matched filter is not about recovering an original waveform.
Scroll down to the match filter recovery of the binary signal
Bret Cahill
Bret Cahill
Apr 6, 2012, 2:21:02 PM4/6/12
to
> What you've been describing isn't a
> matched filtering process,
It works just fine on Excel, the recovery of the original signal
> matched filtering process,
waveform as well as the method outlined in the OP to recover the
original amplitude. So unless there is some other filter out there
that does what I'm doing, it's time to claim this puppy in a method
patent.
EXTRA CREDIT:
If you multiply the FFTs of a noisy signal and its kernel on the SPICE
electronics simulator -- this is faster and easier than Excel -- what
would this represent?
a. a match filter
b. a wiener filter
c. a kalman filter
Bret Cahill
robert bristow-johnson
Apr 6, 2012, 2:32:09 PM4/6/12
to
signal, rather than determining if a 0 or 1 had been transmitted, then
you're mistaken.
also, Wikipedia should not be taken as authoritative, even though
sometimes it can be useful. and that wikipedia article is horrendously
written (as well as the article on instantaneous phase/frequency). i
might suggest finding a good communications textbook. maybe one by A
Bruce Carlson or another by Haykin.
finally, you need to learn to be nice when asking for help. if you
know-it-all, then you don't need any advice or insight from me.
g'bye.
Eric Jacobsen
Apr 6, 2012, 2:37:44 PM4/6/12
to
On Fri, 6 Apr 2012 11:21:02 -0700 (PDT), Bret Cahill
<BretC...@peoplepc.com> wrote:
>> What you've been describing isn't a
>> matched filtering process,
>
>It works just fine on Excel, the recovery of the original signal
>waveform as well as the method outlined in the OP to recover the
>original amplitude. So unless there is some other filter out there
>that does what I'm doing, it's time to claim this puppy in a method
>patent.
Knock yourself out, but if you call it a matched filter or describe it
<BretC...@peoplepc.com> wrote:
>> What you've been describing isn't a
>> matched filtering process,
>
>It works just fine on Excel, the recovery of the original signal
>waveform as well as the method outlined in the OP to recover the
>original amplitude. So unless there is some other filter out there
>that does what I'm doing, it's time to claim this puppy in a method
>patent.
as a matched filtering process you'll just be confusing people.
>EXTRA CREDIT:
>
>If you multiply the FFTs of a noisy signal and its kernel on the SPICE
>electronics simulator -- this is faster and easier than Excel -- what
>would this represent?
>
>a. a match filter
>
>b. a wiener filter
>
>c. a kalman filter
>
>
>Bret Cahill
Bret Cahill
Apr 6, 2012, 3:03:28 PM4/6/12
to
> >>>>>> Do you mean matched filtering or is "match" filtering something
> >>>>>> different?
>
> >>>>> North filter.
>
> >>>> Far more commonly known as a "matched filter",
>
> >>> How long did it take you to figure that out?
>
> >>>> which requires no
> >>>> special kind of convolution, only the ordinary, mundane kind. There's
> >>>> no deconvolution involved.
>
> >>> How is the original waveform recovered with a North Filter?
>
> >> Bret, can you stick with the common semantic? before today, i have
> >> never heard of a "North filter".
>
> > Is anyone still whining about "match filter" or "matched filter", both
> > of which google up tens of thousands of on point hits in less time
> > than it takes to play trifling word games.
>
> >> i am assuming from Eric's response and
> >> yours, that it is synonymous with "matched filter".
>
> >> a matched filter is not about recovering an original waveform.
>
> >http://en.wikipedia.org/wiki/Matched_filter
>
> > Scroll down to the match filter recovery of the binary signal
>
> it's about detection.
In that example the "detection" of the signal requires recovering the
> >>>>>> different?
>
> >>>>> North filter.
>
> >>>> Far more commonly known as a "matched filter",
>
> >>> How long did it take you to figure that out?
>
> >>>> which requires no
> >>>> special kind of convolution, only the ordinary, mundane kind. There's
> >>>> no deconvolution involved.
>
> >>> How is the original waveform recovered with a North Filter?
>
> >> Bret, can you stick with the common semantic? before today, i have
> >> never heard of a "North filter".
>
> > Is anyone still whining about "match filter" or "matched filter", both
> > of which google up tens of thousands of on point hits in less time
> > than it takes to play trifling word games.
>
> >> i am assuming from Eric's response and
> >> yours, that it is synonymous with "matched filter".
>
> >> a matched filter is not about recovering an original waveform.
>
> >http://en.wikipedia.org/wiki/Matched_filter
>
> > Scroll down to the match filter recovery of the binary signal
>
> it's about detection.
original waveform.
> if you think that example is about producing a
> signal, rather than determining if a 0 or 1 had been transmitted, then
> you're mistaken.
look like _that_?
> also, Wikipedia should not be taken as authoritative
Maybe you think the author put that nonsense in because he wanted to
help out anyone trying to mislead on how match filtering can be used
for wave form recovery?
Bret Cahill
Bret Cahill
Apr 6, 2012, 3:13:09 PM4/6/12
to
> >> What you've been describing isn't a
> >> matched filtering process,
> >It works just fine on Excel, the recovery of the original signal
> >waveform as well as the method outlined in the OP to recover the
> >original amplitude. So unless there is some other filter out there
> >that does what I'm doing, it's time to claim this puppy in a method
> >patent.
> Knock yourself out, but if you call it a matched filter or describe it
> as a matched filtering process you'll just be confusing people.
Humor.
> >> matched filtering process,
> >It works just fine on Excel, the recovery of the original signal
> >waveform as well as the method outlined in the OP to recover the
> >original amplitude. So unless there is some other filter out there
> >that does what I'm doing, it's time to claim this puppy in a method
> >patent.
> Knock yourself out, but if you call it a matched filter or describe it
> as a matched filtering process you'll just be confusing people.
I'd love to flatter myself into thinking I discovered a new signal
processing filter. After all, I've been saying all along that there
is low hanging fruit all over the place in engineering applications.
But the reality is fundamentally new filters from a mathematics POV
are very rare. They figured out everything related to convolutions in
signal processing decades ago.
> >EXTRA CREDIT:
>
> >If you multiply the FFTs of a noisy signal and its kernel on the SPICE
> >electronics simulator -- this is faster and easier than Excel -- what
> >would this represent?
>
> >a. a match filter
>
> >b. a wiener filter
>
> >c. a kalman filter
>
> >Bret Cahill
>
> As so many others have said, you have a problem communicating.
above?
Bret Cahill
Eric Jacobsen
Apr 6, 2012, 3:54:34 PM4/6/12
to
On Fri, 6 Apr 2012 12:13:09 -0700 (PDT), Bret Cahill
<BretC...@peoplepc.com> wrote:
>> >> What you've been describing isn't a
>> >> matched filtering process,
>
>> >It works just fine on Excel, the recovery of the original signal
>> >waveform as well as the method outlined in the OP to recover the
>> >original amplitude. =A0So unless there is some other filter out there
<BretC...@peoplepc.com> wrote:
>> >> What you've been describing isn't a
>> >> matched filtering process,
>
>> >It works just fine on Excel, the recovery of the original signal
>> >waveform as well as the method outlined in the OP to recover the
>> >that does what I'm doing, it's time to claim this puppy in a method
>> >patent.
>
>> Knock yourself out, but if you call it a matched filter or describe it
>> as a matched filtering process you'll just be confusing people.
>
>Humor.
>
>I'd love to flatter myself into thinking I discovered a new signal
>processing filter. After all, I've been saying all along that there
>is low hanging fruit all over the place in engineering applications.
>
>But the reality is fundamentally new filters from a mathematics POV
>are very rare. They figured out everything related to convolutions in
>signal processing decades ago.
I'm sure nobody else here realizes that.
>> >patent.
>
>> Knock yourself out, but if you call it a matched filter or describe it
>> as a matched filtering process you'll just be confusing people.
>
>Humor.
>
>I'd love to flatter myself into thinking I discovered a new signal
>processing filter. After all, I've been saying all along that there
>is low hanging fruit all over the place in engineering applications.
>
>But the reality is fundamentally new filters from a mathematics POV
>are very rare. They figured out everything related to convolutions in
>signal processing decades ago.
>> >EXTRA CREDIT:
>>
>> >If you multiply the FFTs of a noisy signal and its kernel on the SPICE
>> >electronics simulator -- this is faster and easier than Excel -- what
>> >would this represent?
>>
>>
>> >b. =A0a wiener filter
>>
>> >c. =A0a kalman filter
>>
>> >Bret Cahill
>>
>> As so many others have said, you have a problem communicating.
>
>Can you think of more than one interpretation of what was posted
>above?
An infinite number, depending on what you mean by "its kernel".
>> >Bret Cahill
>>
>> As so many others have said, you have a problem communicating.
>
>Can you think of more than one interpretation of what was posted
>above?
Eric Jacobsen
Apr 6, 2012, 3:57:18 PM4/6/12
to
On Fri, 6 Apr 2012 12:03:28 -0700 (PDT), Bret Cahill
<BretC...@peoplepc.com> wrote:
>> >>>>>> Do you mean matched filtering or is "match" filtering something
>> >>>>>> different?
>>
>> >>>>> North filter.
>>
>> >>>> Far more commonly known as a "matched filter",
>>
>> >>> How long did it take you to figure that out?
>>
>> >>>> which requires no
>> >>>> special kind of convolution, only the ordinary, mundane kind. =A0The=
<BretC...@peoplepc.com> wrote:
>> >>>>>> Do you mean matched filtering or is "match" filtering something
>> >>>>>> different?
>>
>> >>>>> North filter.
>>
>> >>>> Far more commonly known as a "matched filter",
>>
>> >>> How long did it take you to figure that out?
>>
>> >>>> which requires no
>re's
>> >>>> no deconvolution involved.
>>
>> >>> How is the original waveform recovered with a North Filter?
>>
>> >> Bret, can you stick with the common semantic? =A0before today, i have
>> >>>> no deconvolution involved.
>>
>> >>> How is the original waveform recovered with a North Filter?
>>
>> >> never heard of a "North filter".
>>
>> > Is anyone still whining about "match filter" or "matched filter", both
>> > of which google up tens of thousands of on point hits in less time
>> > than it takes to play trifling word games.
>>
>> >> i am assuming from Eric's response and
>> >> yours, that it is synonymous with "matched filter".
>>
>> >> a matched filter is not about recovering an original waveform.
>>
>> >http://en.wikipedia.org/wiki/Matched_filter
>>
>> > Scroll down to the match filter recovery of the binary signal
>>
>> it's about detection.
>
>In that example the "detection" of the signal requires recovering the
>original waveform.
You're calling it a waveform, other people may call it the modulated
>>
>> > Is anyone still whining about "match filter" or "matched filter", both
>> > of which google up tens of thousands of on point hits in less time
>> > than it takes to play trifling word games.
>>
>> >> i am assuming from Eric's response and
>> >> yours, that it is synonymous with "matched filter".
>>
>> >> a matched filter is not about recovering an original waveform.
>>
>> >http://en.wikipedia.org/wiki/Matched_filter
>>
>> > Scroll down to the match filter recovery of the binary signal
>>
>> it's about detection.
>
>In that example the "detection" of the signal requires recovering the
>original waveform.
data, which can be from a symbol alphabet other than binary.
You need to be clear what you mean if you want to be understood.
>> if you think that example is about producing a
>> signal, rather than determining if a 0 or 1 had been transmitted, then
>> you're mistaken.
>
>You think any convolution of a square wave form with its kernel could
>look like _that_?
>> also, Wikipedia should not be taken as authoritative
>
>You think an authoritative article is necessary for an _example_?
>
>Maybe you think the author put that nonsense in because he wanted to
>help out anyone trying to mislead on how match filtering can be used
>for wave form recovery?
Bret Cahill
Apr 6, 2012, 5:11:55 PM4/6/12
to
What's your point?
. . .
> >> also, Wikipedia should not be taken as authoritative
>
> >You think an authoritative article is necessary for an _example_?
>
> >Maybe you think the author put that nonsense in because he wanted to
> >help out anyone trying to mislead on how match filtering can be used
> >for wave form recovery?
>
> Again, you miss the point by quite a long ways.
The link below shows another example of match filtering, not merely
detecting, but recovering the original waveform before the noise was
added:
http://www.complextoreal.com/chapters/mft.pdf
It may help explain why you are having trouble with the term
"match[ed] filter."
Bret Cahill
Eric Jacobsen
Apr 6, 2012, 5:58:07 PM4/6/12
to
when you throw phrases around like "recover the original waveform" in
a context where it isn't typically used. Matched filtering is not
typically used to recover a waveform, although it can be and that way
of looking at it isn't necessarily wrong, but it's up to you to make
clear what you mean if you want people to understand you.
>. . .
>
>> >> also, Wikipedia should not be taken as authoritative
>>
>> >You think an authoritative article is necessary for an _example_?
>>
>> >Maybe you think the author put that nonsense in because he wanted to
>> >help out anyone trying to mislead on how match filtering can be used
>> >for wave form recovery?
>>
>> Again, you miss the point by quite a long ways.
>
>The communication problem is on _your_ end.
doing DSP and matched filtering applications professionally for a few
decades. I'd like to learn about your insights if you have any, but
so far you haven't explained anything new about matched filtering that
makes any sense, and you seem to have some significant misconceptions.
I await your clarifications using explanations or terminology that can
be understood by an experienced practitioner.
>The link below shows another example of match filtering, not merely
>detecting, but recovering the original waveform before the noise was
>added:
>
>http://www.complextoreal.com/chapters/mft.pdf
>
>It may help explain why you are having trouble with the term
>"match[ed] filter."
could help clarify what you mean, and now you point to a document
which properly explains the usual concepts yet doesn't clarify
anything about what you've been describing regarding deconvolution and
frequency-domain processing, or the use of terms as you've used them.
Charan's tutorials are pretty well known and respected for
introductory treatments to topics, which he's good at. I can't figure
out what you're talking about, though, and you seem unable to clarify
it and unwilling to address the clarifying points that have been
raised by others.
Bret Cahill
Apr 7, 2012, 12:38:29 PM4/7/12
to
form?"
Are there _any_ instances of _any_ convolution looking the same as the
original signal?
. . .
> >> Again, you miss the point by quite a long ways.
>
> >The communication problem is on _your_ end.
>
> You're right, I still have no idea what you're talking about,
wave form with match filtering.
> despite
> doing DSP and matched filtering applications professionally for a few
> decades.
it is used to recover the original signal's wave form.
> I'd like to learn about your insights if you have any,
can be used for _any_ kind of filtering that requires a reference,
i.e., PSD, match, weiner, etc. This reference might have some
academic interest in that it is constructed from the noisy signals
themselves. I doubt it but there may be some way to
But the hope that, other than maybe the modeling of match filtering on
Excel, anything mathematically new is going on here is almost
inconceivable.
There are endless opportunitie$ in engineering applications but trying
to come up with something new in math is like invading Afghanistan.
There's no money in it and you are up against the greatest fighters on
earth.
Math Tree was stripped bare a long time ago, not just of fruit, but of
bark, branches, the trunk and root system are all gone. All that's
left is a crater.
That's why you should never bother to google to see if you've ever
discovered anything new in math. It's not worth the time. You know
it was done decades or centuries ago so why bother?
Bret Cahill
Bret Cahill
Apr 9, 2012, 6:50:10 PM4/9/12
to
> Instead of just taking the convolution of a signal with the reference
> or kernel once and then taking the square root in the frequency
> domain, why not match filter with the reference again and again to
> recover the original clean signal?
>
> Well, you _can_ do that if a wave form is all you want, but you can
> get that from the ref. alone _anyway_.
>
> If you want to keep the original magnitude of the signal, however, and
> if the kernel is a different magnitude than the original signal, then
> you will only be working your way back to the kernel with each
> iteration.
>
> The solution here is to convolve the signal with the kernel and then
> the kernel with the kernel and then comparing the magnitudes of those
> two convolutions.
>
> Apply that factor to correct the magnitude of the filtered signal.
After you do that once you're already at your minimum magnitude
> or kernel once and then taking the square root in the frequency
> domain, why not match filter with the reference again and again to
> recover the original clean signal?
>
> Well, you _can_ do that if a wave form is all you want, but you can
> get that from the ref. alone _anyway_.
>
> If you want to keep the original magnitude of the signal, however, and
> if the kernel is a different magnitude than the original signal, then
> you will only be working your way back to the kernel with each
> iteration.
>
> The solution here is to convolve the signal with the kernel and then
> the kernel with the kernel and then comparing the magnitudes of those
> two convolutions.
>
> Apply that factor to correct the magnitude of the filtered signal.
error. Repeating the match filtering more times may make it look more
like the original signal but it will not do anything to change the
magnitude and improve accuracy.
Bret Cahill
Fred Marshall
Apr 9, 2012, 11:06:58 PM4/9/12
to
spike output. (We used to dream about doing this with a pseudo random
sequence. Trouble was, the SNR for the output in a limited bandwidth
was lousy.) Anyway, then convolve that imagined ideal spike with the
original waveform and voila! the original waveform scaled.
But, as someone already asked: "what's the benefit of getting out what
you already know?" Just take the amplitude of the spike.
Fred
Bret Cahill
Apr 10, 2012, 2:19:47 AM4/10/12
to
> >> Instead of just taking the convolution of a signal with the reference
> >> or kernel once and then taking the square root in the frequency
> >> domain, why not match filter with the reference again and again to
> >> recover the original clean signal?
> >> Well, you _can_ do that if a wave form is all you want, but you can
> >> get that from the ref. alone _anyway_.
> >> If you want to keep the original magnitude of the signal, however, and
> >> if the kernel is a different magnitude than the original signal, then
> >> you will only be working your way back to the kernel with each
> >> iteration.
> >> The solution here is to convolve the signal with the kernel and then
> >> the kernel with the kernel and then comparing the magnitudes of those
> >> two convolutions.
> >> Apply that factor to correct the magnitude of the filtered signal.
> > After you do that once you're already at your minimum magnitude
> > error. Repeating the match filtering more times may make it look more
> > like the original signal but it will not do anything to change the
> > magnitude and improve accuracy.
> >> or kernel once and then taking the square root in the frequency
> >> domain, why not match filter with the reference again and again to
> >> recover the original clean signal?
> >> Well, you _can_ do that if a wave form is all you want, but you can
> >> get that from the ref. alone _anyway_.
> >> If you want to keep the original magnitude of the signal, however, and
> >> if the kernel is a different magnitude than the original signal, then
> >> you will only be working your way back to the kernel with each
> >> iteration.
> >> The solution here is to convolve the signal with the kernel and then
> >> the kernel with the kernel and then comparing the magnitudes of those
> >> two convolutions.
> >> Apply that factor to correct the magnitude of the filtered signal.
> > After you do that once you're already at your minimum magnitude
> > error. Repeating the match filtering more times may make it look more
> > like the original signal but it will not do anything to change the
> > magnitude and improve accuracy.
> Now that I think about it, the "perfect" matched filter would generate a
> spike output.
Depends on what you want from the matched filter.
> spike output.
The low pass filtering effect is nice because most of the noise is
about one decade higher than the signal in my application.
> (We used to dream about doing this with a pseudo random
> sequence. Trouble was, the SNR for the output in a limited bandwidth
> was lousy.) Anyway, then convolve that imagined ideal spike with the
> original waveform and voila! the original waveform scaled.
> But, as someone already asked: "what's the benefit of getting out what
> you already know?" Just take the amplitude of the spike.
filtering when inverse filtering isn't necessary?
Supposing inverse filtering _is_ necessary. Is Wiener the way to go?
Bret Cahill
fatalist
Apr 10, 2012, 10:18:24 AM4/10/12
to
On Apr 3, 1:25 am, Bret Cahill <BretCah...@peoplepc.com> wrote:
> Instead of just taking the convolution of a signal with the reference
> or kernel once and then taking the square root in the frequency
> domain, why not match filter with the reference again and again to
> recover the original clean signal?
>
> Well, you _can_ do that if a wave form is all you want, but you can
> get that from the ref. alone _anyway_.
>
> If you want to keep the original magnitude of the signal, however, and
> if the kernel is a different magnitude than the original signal, then
> you will only be working your way back to the kernel with each
> iteration.
>
> The solution here is to convolve the signal with the kernel and then
> the kernel with the kernel and then comparing the magnitudes of those
> two convolutions.
>
> Apply that factor to correct the magnitude of the filtered signal.
>
> Bret Cahill
> Instead of just taking the convolution of a signal with the reference
> or kernel once and then taking the square root in the frequency
> domain, why not match filter with the reference again and again to
> recover the original clean signal?
>
> Well, you _can_ do that if a wave form is all you want, but you can
> get that from the ref. alone _anyway_.
>
> If you want to keep the original magnitude of the signal, however, and
> if the kernel is a different magnitude than the original signal, then
> you will only be working your way back to the kernel with each
> iteration.
>
> The solution here is to convolve the signal with the kernel and then
> the kernel with the kernel and then comparing the magnitudes of those
> two convolutions.
>
> Apply that factor to correct the magnitude of the filtered signal.
>
"The matched filter is the optimal linear filter for maximizing the
signal to noise ratio (SNR) in the presence of additive stochastic
noise" (http://en.wikipedia.org/wiki/Matched_filter)
LINEAR is the key word here
If you remove "linear" from the above sentence then "matched filter"
is not even close to being "optimal" in any practical sense
Go figure how to build an optimal non-linear "matched filter" if you
can
Good luck
Bret Cahill
Apr 10, 2012, 1:57:59 PM4/10/12
to
> > Instead of just taking the convolution of a signal with the reference
> > or kernel once and then taking the square root in the frequency
> > domain, why not match filter with the reference again and again to
> > recover the original clean signal?
>
> > Well, you _can_ do that if a wave form is all you want, but you can
> > get that from the ref. alone _anyway_.
>
> > If you want to keep the original magnitude of the signal, however, and
> > if the kernel is a different magnitude than the original signal, then
> > you will only be working your way back to the kernel with each
> > iteration.
>
> > The solution here is to convolve the signal with the kernel and then
> > the kernel with the kernel and then comparing the magnitudes of those
> > two convolutions.
>
> > Apply that factor to correct the magnitude of the filtered signal.
>
> > Bret Cahill
>
> "The matched filter is the optimal linear filter for maximizing the
> signal to noise ratio (SNR) in the presence of additive stochastic
> noise" (http://en.wikipedia.org/wiki/Matched_filter)
>
> LINEAR is the key word here
Linearity was the assumption of the OP as nothing new was being done
> > or kernel once and then taking the square root in the frequency
> > domain, why not match filter with the reference again and again to
> > recover the original clean signal?
>
> > Well, you _can_ do that if a wave form is all you want, but you can
> > get that from the ref. alone _anyway_.
>
> > If you want to keep the original magnitude of the signal, however, and
> > if the kernel is a different magnitude than the original signal, then
> > you will only be working your way back to the kernel with each
> > iteration.
>
> > The solution here is to convolve the signal with the kernel and then
> > the kernel with the kernel and then comparing the magnitudes of those
> > two convolutions.
>
> > Apply that factor to correct the magnitude of the filtered signal.
>
> > Bret Cahill
>
> "The matched filter is the optimal linear filter for maximizing the
> signal to noise ratio (SNR) in the presence of additive stochastic
> noise" (http://en.wikipedia.org/wiki/Matched_filter)
>
> LINEAR is the key word here
until the deconvolution.
The reference is just the noise free signal times a scalar so all you
are really doing when you take the product of the template and noise
free signal in the frequency domain is taking the square of the
reference or signal times a scalar.
So taking the square root in the frequency domain is the reverse
operation of multiplying the noise free signal by the reference in the
frequency domain.
That's how you deconvolve what some consider the final output of a
match filter.
This was first done a few months ago on Excel at http://www.bretcahill.com
The original goal was to filter the signal for amplitude. If the
convolution works best, use the convolution. If recovering the
filtered signal by usig the deconvolution works better then use that.
Bret Cahill
Fred Marshall
Apr 10, 2012, 4:07:44 PM4/10/12
to
On 4/9/2012 11:19 PM, Bret Cahill wrote:
> Depends on what you want from the matched filter
I want to do the classical thing. I don't know of any other using that
> Depends on what you want from the matched filter
terminology. Terminology remains an issue in this thread.
My objective would be to find the time delay or advance which maximizes
the signal to noise ratio *of the inner product of two signal vectors or
sequences*.
Note that the inner product can have a slowly changing value as a
function of the length of the replica. That is, for a square pulse of
length N samples, the output will change by 1/N times the difference
between the first and last samples. Maybe that's a case where you
mention "lowpass".
Anyway, the output at each sample time is a single sample.
To do something else is to do something else which has a different name.
For example, I can imagine this:
1) Define a signal
2) Based on that signal, define a matched filter.
3) Run a perturbed version of the original signal through the matched
filter.
4) Examine the output for a peak which will estimate the time delay with
highest possible SNR of the *PEAK* output.
So that's what we mean by "matched filtering".
But, what if you do 1,2 and 3 but not 4?
What if you substitute for 4 something like this:
4) Examine the output for a shape that looks like the replica.
What do you call that? It's not a matched filter in the classical sense.
How is "examine" implemented in this process?
What if "examine" is implemented by doing yet another correlation? I
think maybe that's what you had in mind.
Then, doing another correlation, in view of the objective, what do you
look for? A peak no doubt, right?
This gives rise to a few questions:
Q1: Why would one think that the output of a classical matched filter
would look like the replica .. even in the noisless or unperturbed case?
A1: One would not expect that to be the case.
Q2: Given Q1/A1 then what would one expect?
A2: Use a square pulse as an example.
The matched filter output has a triangular wave shape with a peak
where the square pulses coincide. The same happens with a square pulse
shaped sinusoid. The output envelope is triangular.
Re correlating this with another square pulse does nothing very good to
restore the original square pulse shape.
If this post isn't helpful then maybe it will suggest ways that the
discussion can be cleaned up so we understand what's being:
- asked?
- claimed?
- asserted?
what???? It is very much not clear Bret.
Fred
Bret Cahill
Apr 10, 2012, 5:36:09 PM4/10/12
to
> > Depends on what you want from the matched filter
> I want to do the classical thing.
But supposing matched filtering could be adapted for another use where
> I want to do the classical thing.
it was better than another popular filter, say, a Wiener filter?
> I don't know of any other using that
> terminology. Terminology remains an issue in this thread.
an expression that means, "the exact right word." I generally agree
but everything seems so straightforward here any formality whatsoever
seems silly.
Just follow the steps in Excel in the optics thread to understand how
it's being done but a good mathematician should be able to figure it
out just from the 2 or 3 paragraphs at http://www.bretcahill.com
If anyone wants I'll email the Excel program. The colors of the cells
are the same as those in the graphs.
. . .
> To do something else is to do something else which has a different name.
> For example, I can imagine this:
> 1) Define a signal
> 2) Based on that signal, define a matched filter.
> 3) Run a perturbed version of the original signal through the matched
> filter.
> 4) Examine the output for a peak which will estimate the time delay with
> highest possible SNR of the *PEAK* output.
> So that's what we mean by "matched filtering".
signal? Except for _maybe_ some exponential curve, there are no
convolutions that look like the original signal.
So, if your goal is to recover the original signal without nearly as
much noise, why not use all the information in the template /
reference? Why not use all the phase angle information as well as the
PSD that is provided by matched filtering?
There's more to life than radar and other signal detection solutions.
> But, what if you do 1,2 and 3 but not 4?
> What if you substitute for 4 something like this:
> 4) Examine the output for a shape that looks like the replica.
your step 3.
This is easy because all you need to do is IMSQRT the convolution in
the frequency domain.
This recovers the original signal's wave form minus a lot of noise.
> What do you call that? It's not a matched filter in the classical sense.
I was so sure they had been doing this for decades I never even
bothered to google it.
> How is "examine" implemented in this process?
> What if "examine" is implemented by doing yet another correlation? I
> think maybe that's what you had in mind.
> Then, doing another correlation, in view of the objective, what do you
> look for? A peak no doubt, right?
>
> This gives rise to a few questions:
>
> Q1: Why would one think that the output of a classical matched filter
> would look like the replica .. even in the noisless or unperturbed case?
> A1: One would not expect that to be the case.
>
> Q2: Given Q1/A1 then what would one expect?
>
> A2: Use a square pulse as an example.
> The matched filter output has a triangular wave shape with a peak
> where the square pulses coincide.
triangles back to the original square pulses?
Just take the square root in the frequency domain then inverse FFT.
A $250 netbook will easily do the dozen or so 512 sample convolutions
in a couple of seconds.
> The same happens with a square pulse
> shaped sinusoid. The output envelope is triangular.
> Re correlating this with another square pulse does nothing very good to
> restore the original square pulse shape.
> If this post isn't helpful then maybe it will suggest ways that the
> discussion can be cleaned up so we understand what's being:
> - asked?
> - claimed?
> - asserted?
> what???? It is very much not clear Bret.
Bret Cahill
Eric Jacobsen
Apr 10, 2012, 6:12:45 PM4/10/12
to
On Tue, 10 Apr 2012 14:36:09 -0700 (PDT), Bret Cahill
<BretC...@peoplepc.com> wrote:
>> > Depends on what you want from the matched filter
>
>> I want to do the classical thing.
>
>But supposing matched filtering could be adapted for another use where
>it was better than another popular filter, say, a Wiener filter?
Supposing you can drive a screw into a piece of wood faster with a
<BretC...@peoplepc.com> wrote:
>> > Depends on what you want from the matched filter
>
>> I want to do the classical thing.
>
>But supposing matched filtering could be adapted for another use where
>it was better than another popular filter, say, a Wiener filter?
hammer than a screwdriver, does that mean it's a better method?
Fred Marshall
Apr 10, 2012, 7:16:05 PM4/10/12
to
On 4/10/2012 2:36 PM, Bret Cahill wrote:
>> If this post isn't helpful then maybe it will suggest ways that the
>> > discussion can be cleaned up so we understand what's being:
>> > - asked?
>> > - claimed?
>> > - asserted?
>> > what???? It is very much not clear Bret.
> There's very little to work with to clarify.
I was suggesting that my post may *not* be helpful which would suggest
>> If this post isn't helpful then maybe it will suggest ways that the
>> > discussion can be cleaned up so we understand what's being:
>> > - asked?
>> > - claimed?
>> > - asserted?
>> > what???? It is very much not clear Bret.
> There's very little to work with to clarify.
in turn that you clarify whether you are asking, claiming or asserting.
I get it now. You're asserting. I'm not further interested as I
clearly cannot help you assert.
Fred
Bret Cahill
Apr 11, 2012, 12:26:27 AM4/11/12
to
didn't do some asserting myself.
Asserting doesn't preclude the asking of questions either.
Bret Cahill
robert bristow-johnson
Apr 11, 2012, 1:25:53 AM4/11/12
to
On 4/11/12 12:26 AM, Bret Cahill wrote:
>
> _Everyone_ needs to start asserting more.
sometimes society (the rest of us) need some people to assert less and
>
> _Everyone_ needs to start asserting more.
listen more. the "needs" of the individual to assert are secondary.
> I'd be a hypocrite if I didn't do some asserting myself.
> Asserting doesn't preclude the asking of questions either.
answer.
Bret Cahill
Apr 11, 2012, 1:57:59 AM4/11/12
to
> > _Everyone_ needs to start asserting more.
>
> sometimes society (the rest of us) need some people to assert less and
> listen more.
That was never true, even _before_ everything was online.
>
> sometimes society (the rest of us) need some people to assert less and
> listen more.
Bret Cahill
"Anything less than genius just won't do."
-- English woman commentator on BBC (2006)
Randy Yates
Apr 11, 2012, 3:09:18 AM4/11/12
to
eric.j...@ieee.org (Eric Jacobsen) writes:
> On Tue, 10 Apr 2012 14:36:09 -0700 (PDT), Bret Cahill
> <BretC...@peoplepc.com> wrote:
>
>>> > Depends on what you want from the matched filter
>>
>>> I want to do the classical thing.
>>
>>But supposing matched filtering could be adapted for another use where
>>it was better than another popular filter, say, a Wiener filter?
>
> Supposing you can drive a screw into a piece of wood faster with a
> hammer than a screwdriver, does that mean it's a better method?
Eric, have you ever considered philosophy as a career? :)
> On Tue, 10 Apr 2012 14:36:09 -0700 (PDT), Bret Cahill
> <BretC...@peoplepc.com> wrote:
>
>>> > Depends on what you want from the matched filter
>>
>>> I want to do the classical thing.
>>
>>But supposing matched filtering could be adapted for another use where
>>it was better than another popular filter, say, a Wiener filter?
>
> Supposing you can drive a screw into a piece of wood faster with a
> hammer than a screwdriver, does that mean it's a better method?
--
Randy Yates
DSP/Firmware Engineer
919-577-9882 (H)
919-720-2916 (C)
Bret Cahill
Apr 11, 2012, 10:53:43 AM4/11/12
to
> >>> > Depends on what you want from the matched filter
>
> >>> I want to do the classical thing.
>
> >>But supposing matched filtering could be adapted for another use where
> >>it was better than another popular filter, say, a Wiener filter?
>
> > Supposing you can drive a screw into a piece of wood faster with a
> > hammer than a screwdriver, does that mean it's a better method?
>
> Eric, have you ever considered philosophy as a career? :)
He dodged the issue so he might work out as a GOP political
>
> >>> I want to do the classical thing.
>
> >>But supposing matched filtering could be adapted for another use where
> >>it was better than another popular filter, say, a Wiener filter?
>
> > Supposing you can drive a screw into a piece of wood faster with a
> > hammer than a screwdriver, does that mean it's a better method?
>
> Eric, have you ever considered philosophy as a career? :)
"philosopher."
Match filtering, or, if the additional deconvolution step is new,
"Matched Filtering/Magnitude Recovery" utilizes all of the phase angle
info as well as the PSD info and is, therefore, better than any other
filter as far as recovering the magnitude of the original signal.
Matched filtering/signal recovery also seems to be a better way to
recover square waves out of noise than screwing around with triangle
shaped convolutions.
Bret Cahill
Eric Jacobsen
Apr 11, 2012, 11:37:02 AM4/11/12
to
On Wed, 11 Apr 2012 07:53:43 -0700 (PDT), Bret Cahill
<BretC...@peoplepc.com> wrote:
>> >>> > Depends on what you want from the matched filter
>>
>> >>> I want to do the classical thing.
>>
>> >>But supposing matched filtering could be adapted for another use where
>> >>it was better than another popular filter, say, a Wiener filter?
>>
>> > Supposing you can drive a screw into a piece of wood faster with a
>> > hammer than a screwdriver, =A0does that mean it's a better method?
<BretC...@peoplepc.com> wrote:
>> >>> > Depends on what you want from the matched filter
>>
>> >>> I want to do the classical thing.
>>
>> >>But supposing matched filtering could be adapted for another use where
>> >>it was better than another popular filter, say, a Wiener filter?
>>
>> > Supposing you can drive a screw into a piece of wood faster with a
>>
>> Eric, have you ever considered philosophy as a career? :)
>
>He dodged the issue so he might work out as a GOP political
>"philosopher."
>
>Match filtering, or, if the additional deconvolution step is new,
>"Matched Filtering/Magnitude Recovery" utilizes all of the phase angle
>info as well as the PSD info and is, therefore, better than any other
>filter as far as recovering the magnitude of the original signal.
"Matched" filtering can use all of the phase info, too, and very often
>> Eric, have you ever considered philosophy as a career? :)
>
>He dodged the issue so he might work out as a GOP political
>"philosopher."
>
>Match filtering, or, if the additional deconvolution step is new,
>"Matched Filtering/Magnitude Recovery" utilizes all of the phase angle
>info as well as the PSD info and is, therefore, better than any other
>filter as far as recovering the magnitude of the original signal.
does. You still seem to have a lot of misconceptions about what
matched filtering is, what it does, and how it does it.
What you've been describing is not matched filtering.
>Matched filtering/signal recovery also seems to be a better way to
>recover square waves out of noise than screwing around with triangle
>shaped convolutions.
filtering against a square-wave input. If you get anything else out,
it wasn't a matched filtering process. Your whole notion of
"recovering the original waveform" is NOT a matched filtering process
or objective in any usual sense. This has been pointed out to you
multiple times.
You're the one that cited Charan's tutorial page on matched filtering.
You should take the time to actually understand it.
Bret Cahill
Apr 11, 2012, 1:42:48 PM4/11/12
to
> >> >>> > Depends on what you want from the matched filter
>
> >> >>> I want to do the classical thing.
>
> >> >>But supposing matched filtering could be adapted for another use where
> >> >>it was better than another popular filter, say, a Wiener filter?
>
> >> > Supposing you can drive a screw into a piece of wood faster with a
> >> > hammer than a screwdriver, =A0does that mean it's a better method?
>
> >> Eric, have you ever considered philosophy as a career? :)
>
> >He dodged the issue so he might work out as a GOP political
> >"philosopher."
>
> >Match filtering, or, if the additional deconvolution step is new,
> >"Matched Filtering/Magnitude Recovery" utilizes all of the phase angle
> >info as well as the PSD info and is, therefore, better than any other
> >filter as far as recovering the magnitude of the original signal.
>
> "Matched" filtering can use all of the phase info, too, and very often
> does.
That's why I've been calling it "match filtering."
>
> >> >>> I want to do the classical thing.
>
> >> >>But supposing matched filtering could be adapted for another use where
> >> >>it was better than another popular filter, say, a Wiener filter?
>
> >> > Supposing you can drive a screw into a piece of wood faster with a
> >> > hammer than a screwdriver, =A0does that mean it's a better method?
>
> >> Eric, have you ever considered philosophy as a career? :)
>
> >He dodged the issue so he might work out as a GOP political
> >"philosopher."
>
> >Match filtering, or, if the additional deconvolution step is new,
> >"Matched Filtering/Magnitude Recovery" utilizes all of the phase angle
> >info as well as the PSD info and is, therefore, better than any other
> >filter as far as recovering the magnitude of the original signal.
>
> "Matched" filtering can use all of the phase info, too, and very often
> does.
> You still seem to have a lot of misconceptions about what
> matched filtering is, what it does, and how it does it.
> What you've been describing is not matched filtering.
info as well as power spectrum info I'd probably use it.
If not then I'll stick to using the only term where that holds.
> >Matched filtering/signal recovery also seems to be a better way to
> >recover square waves out of noise than screwing around with triangle
> >shaped convolutions.
>
> The triangle-shaped convolution output *IS* the result of matched
> filtering against a square-wave input.
> If you get anything else out,
> it wasn't a matched filtering process.
get the final deconvolution.
> Your whole notion of
> "recovering the original waveform" is NOT a matched filtering process
> or objective in any usual sense.
Maybe another term reflecting the origins of the filter would be
acceptable, i.e., "Match Filtering Plus" or "Matched Filtering/Signal
Recovery."
Not that I'm humble or anything but the "Bret Filter" doesn't explain
it very well. That's why "North filter" never caught on.
Bret Cahill
Bret Cahill
Apr 15, 2012, 1:52:40 PM4/15/12
to
> > Anyone who says deconvolution isn't complicated demonstrably does not
> > have the first clue about the subject of signal processing. Convolution
> > is easy but deconvolution is usually a very difficult inverse problem.
>
> Well, linear deconvolution isn't complicated,
But, hey, a dunce wanted to sound like he knew something about math.
> but it also often
> gives less than useful results.
And it often gives useful results.
In any event, from the beginning the complete process of recovering
the original signal should have been considered.
Apparently this is something new as well as easy.
Bret Cahill
> > have the first clue about the subject of signal processing. Convolution
> > is easy but deconvolution is usually a very difficult inverse problem.
>
> Well, linear deconvolution isn't complicated,
But, hey, a dunce wanted to sound like he knew something about math.
> but it also often
> gives less than useful results.
And it often gives useful results.
In any event, from the beginning the complete process of recovering
the original signal should have been considered.
Apparently this is something new as well as easy.
Bret Cahill
brent
Apr 15, 2012, 1:59:50 PM4/15/12
to
someone steals your idea...
Bret Cahill
Apr 17, 2012, 1:24:39 PM4/17/12
to
brent
Apr 17, 2012, 2:37:01 PM4/17/12
to
On Apr 17, 1:24 pm, Bret Cahill <Bret_E_Cah...@yahoo.com> wrote:
> > > Anyone who says deconvolution isn't complicated demonstrably does not
> > > have the first clue about the subject of signal processing. Convolution
> > > is easy but deconvolution is usually a very difficult inverse problem.
> > Well, linear deconvolution isn't complicated,
>
> But, hey, a dunce wanted to sound like he knew something about math.
>
We get it, you are a genius, and everyone else are dunces.
> > > Anyone who says deconvolution isn't complicated demonstrably does not
> > > have the first clue about the subject of signal processing. Convolution
> > > is easy but deconvolution is usually a very difficult inverse problem.
> > Well, linear deconvolution isn't complicated,
>
> But, hey, a dunce wanted to sound like he knew something about math.
>
glen herrmannsfeldt
Apr 17, 2012, 3:10:47 PM4/17/12
to
In comp.dsp brent <bule...@columbus.rr.com> wrote:
(snip on deconvolution)
>> But, hey, a dunce wanted to sound like he knew something about math.
http://www.amazon.com/s/ref=nb_sb_noss?url=search-alias%3Dstripbooks&field-keywords=0486453251
It seems that it is available on June 20th.
(Moon day, maybe that isn't a coincidence.)
-- glen
(snip on deconvolution)
>> But, hey, a dunce wanted to sound like he knew something about math.
It seems that it is available on June 20th.
(Moon day, maybe that isn't a coincidence.)
-- glen
Bret Cahill
Apr 17, 2012, 3:56:18 PM4/17/12
to
> >> But, hey, a dunce wanted to sound like he knew something about math.
>
> http://www.amazon.com/s/ref=nb_sb_noss?url=search-alias%3Dstripbooks&...
>
>
> It seems that it is available on June 20th.
Apparently there is no disagreement on the following:
> It seems that it is available on June 20th.
1. What is and has been called "matched filtering" for 7 decades
doesn't recover the original signal, just the convolution.
2. The recovery of the original signal from the output of a matched
filter is easy as all you need to do is take the square root of the
convolution in the frequency domain, in excel IMSQRT of the FFT of the
(output in the time domain).
3. There was never any reason for North not to consider trying to
recover the original signal back in 1943 other than the distraction of
a war.
4. There are many reasons to consider recovering the original signal.
a. the "matched" part of the filtering may be treated as a distinct
step that may be separate from the low pass filtering effect of the
convolution.
b. operations such as squelching the signal when the signal's 1st
derivative exceeds a threshold are easy.
c. it elegantly recovers a square / binary wave with screwing around
with triangles.
d. it's mathematically correct.
Bret Cahill
brent
Apr 17, 2012, 6:06:34 PM4/17/12
to
On Apr 17, 3:56 pm, Bret Cahill <Bret_E_Cah...@yahoo.com> wrote:
>
> Apparently there is no disagreement on the following:
>
Apparently not...
>
> Apparently there is no disagreement on the following:
>
brent
Apr 17, 2012, 6:05:55 PM4/17/12
to
On Apr 17, 3:56 pm, Bret Cahill <Bret_E_Cah...@yahoo.com> wrote:
>
> Apparently there is no disagreement on the following:
Apparently not...
>
> Apparently there is no disagreement on the following:
brent
Apr 17, 2012, 6:13:22 PM4/17/12
to
0 new messages