How can i separate three signals based on frequency and amplitude?

Thanks

abdul khan

If the three signals are in different frequency bands and those bands don't
overlap then you should be able to separate them by filtering without
worrying too much about the amplitude - is this the case?

"Abdul Nasir Khan" <a...@ieee.org> wrote in message
news:9...@posting.google.com...
> How can i separate three signals based on frequency and amplitude?
>
> Thanks
>
> abdul khan

Well..I am very new to DSP area. I would appreciate algorithm in the
form of psedue code.

thanks and sincerely,

abdul nasir khan

"Mike Yarwood" <m...@btopenworld.com> wrote in message
news:<c1nkrg$5dp$1...@titan.btinternet.com>...
> If the three signals are in different frequency bands and those bands don't
> overlap then you should be able to separate them by filtering without
> worrying too much about the amplitude - is this the case?
> 
> "Abdul Nasir Khan" <a...@ieee.org> wrote in message
> news:9...@posting.google.com...
> > How can i separate three signals based on frequency and amplitude?
> >
> > Thanks
> >
> > abdul khan

You're not going to get a psuedo code algorithm for this...
There are just too many ways of going about it.

You are looking to essentially create three seperate parallel bandpass
filters, each one will let through a single signal whilst blocking the
remaining two signals.  Hence from the output of the three bandpass filters
you will have three seperated signals.
We would need more information about the signals in particular to give more
information about whether it's an easy task to achieve this filtering.
You could do this in the tiem domain by using a FIR or IIR filter, or you
could do it in the frequency domain by performing an FFT (or DFT) on the
time domain data then multiplying the frequency bins by some function to
remove the undesired frequencies, then simply perform an IFFT (or IDCT) to
get back into the time domain.

If you perform a google search for this you should get plenty of links to
allow you to perform some research of your own.

"Abdul Nasir Khan" <a...@ieee.org> wrote in message
news:9...@posting.google.com...
> Well..I am very new to DSP area. I would appreciate algorithm in the
> form of psedue code.
>
> thanks and sincerely,
>
> abdul nasir khan
>
> "Mike Yarwood" <m...@btopenworld.com> wrote in message
news:<c1nkrg$5dp$1...@titan.btinternet.com>...
> > If the three signals are in different frequency bands and those bands
don't
> > overlap then you should be able to separate them by filtering without
> > worrying too much about the amplitude - is this the case?
> >
> > "Abdul Nasir Khan" <a...@ieee.org> wrote in message
> > news:9...@posting.google.com...
> > > How can i separate three signals based on frequency and amplitude?
> > >
> > > Thanks
> > >
> > > abdul khan

But certainly, your mail is of very much help to me. I will now study
i) Band Pass Filter
ii) Application FIR, when signal is in time domain
iii) Multiplication of some XYZ function after applying FT on time
domain data.

All of this expercise is only for my self-learning and is not
associated any real project at all. Therefore At the moment I myself
don't know to what kind of signals is it? In fact I will have to
generate my own signals, but how?.

...I am like working in the dark room and trying to look for a lost
needle.

Any help will be appreciated in this context...

Thanks

Abdul N Khan
a.nasir @ ieee org

"Bevan Weiss" <k...@NOSPAMhotmail.com> wrote in message
news:<3YO0c.7372$S...@news.xtra.co.nz>...
> You're not going to get a psuedo code algorithm for this...
> There are just too many ways of going about it.
> 
> You are looking to essentially create three seperate parallel bandpass
> filters, each one will let through a single signal whilst blocking the
> remaining two signals.  Hence from the output of the three bandpass filters
> you will have three seperated signals.
> We would need more information about the signals in particular to give more
> information about whether it's an easy task to achieve this filtering.
> You could do this in the tiem domain by using a FIR or IIR filter, or you
> could do it in the frequency domain by performing an FFT (or DFT) on the
> time domain data then multiplying the frequency bins by some function to
> remove the undesired frequencies, then simply perform an IFFT (or IDCT) to
> get back into the time domain.
> 
> If you perform a google search for this you should get plenty of links to
> allow you to perform some research of your own.
> 
> "Abdul Nasir Khan" <a...@ieee.org> wrote in message
> news:9...@posting.google.com...
> > Well..I am very new to DSP area. I would appreciate algorithm in the
> > form of psedue code.
> >
> > thanks and sincerely,
> >
> > abdul nasir khan
> >
> > "Mike Yarwood" <m...@btopenworld.com> wrote in message
>  news:<c1nkrg$5dp$1...@titan.btinternet.com>...
> > > If the three signals are in different frequency bands and those bands
>  don't
> > > overlap then you should be able to separate them by filtering without
> > > worrying too much about the amplitude - is this the case?
> > >
> > > "Abdul Nasir Khan" <a...@ieee.org> wrote in message
> > > news:9...@posting.google.com...
> > > > How can i separate three signals based on frequency and amplitude?
> > > >
> > > > Thanks
> > > >
> > > > abdul khan

Generating a signal can be the easy part, depending on what kind of
constraints you wish to place on it.

You could generate the data point values on the fly by using some kind of
taylor series, or you could use a look up table with some kind of phase
accumulator (my preferred option if you're generating multiple waveforms at
different frequencies).  You could always use an analog sinusoidal
oscillator and then some ADC to sample the waveform.  Any of these methods
will generate a 'simple' sinusoid.

I think probably the best method would be for you to get a simple
development board with a big player DSP on it, and several channels of ADC
and DAC onboard.  You will also want some signal generators, perhaps you
could make these yourself using some basic circuitry (assuming you're not
doing anything high end just yet).
You can then just play around with inputting several different frequency
signals and trying to seperate them out into their own channels.  See what
kind of effect the frequency spacing between the input signals is, and what
effect it has on the seperated channels.
It would give you a chance to experiment with the different filter methods,
as well as other DSP topics.


"Abdul Nasir Khan" <a...@ieee.org> wrote in message
news:9...@posting.google.com...
> But certainly, your mail is of very much help to me. I will now study
> i) Band Pass Filter
> ii) Application FIR, when signal is in time domain
> iii) Multiplication of some XYZ function after applying FT on time
> domain data.
>
> All of this expercise is only for my self-learning and is not
> associated any real project at all. Therefore At the moment I myself
> don't know to what kind of signals is it? In fact I will have to
> generate my own signals, but how?.
>
> ...I am like working in the dark room and trying to look for a lost
> needle.
>
> Any help will be appreciated in this context...
>
> Thanks
>
> Abdul N Khan
> a.nasir @ ieee org
>
> "Bevan Weiss" <k...@NOSPAMhotmail.com> wrote in message
news:<3YO0c.7372$S...@news.xtra.co.nz>...
> > You're not going to get a psuedo code algorithm for this...
> > There are just too many ways of going about it.
> >
> > You are looking to essentially create three seperate parallel bandpass
> > filters, each one will let through a single signal whilst blocking the
> > remaining two signals.  Hence from the output of the three bandpass
filters
> > you will have three seperated signals.
> > We would need more information about the signals in particular to give
more
> > information about whether it's an easy task to achieve this filtering.
> > You could do this in the tiem domain by using a FIR or IIR filter, or
you
> > could do it in the frequency domain by performing an FFT (or DFT) on the
> > time domain data then multiplying the frequency bins by some function to
> > remove the undesired frequencies, then simply perform an IFFT (or IDCT)
to
> > get back into the time domain.
> >
> > If you perform a google search for this you should get plenty of links
to
> > allow you to perform some research of your own.
> >
> > "Abdul Nasir Khan" <a...@ieee.org> wrote in message
> > news:9...@posting.google.com...
> > > Well..I am very new to DSP area. I would appreciate algorithm in the
> > > form of psedue code.
> > >
> > > thanks and sincerely,
> > >
> > > abdul nasir khan
> > >
> > > "Mike Yarwood" <m...@btopenworld.com> wrote in message
> >  news:<c1nkrg$5dp$1...@titan.btinternet.com>...
> > > > If the three signals are in different frequency bands and those
bands
> >  don't
> > > > overlap then you should be able to separate them by filtering
without
> > > > worrying too much about the amplitude - is this the case?
> > > >
> > > > "Abdul Nasir Khan" <a...@ieee.org> wrote in message
> > > > news:9...@posting.google.com...
> > > > > How can i separate three signals based on frequency and amplitude?
> > > > >
> > > > > Thanks
> > > > >
> > > > > abdul khan

I am working at very basic level.
My ultimate objective is implementing a project for learning that
separates the noises from sound captured.

At the moment I am trying to get myself started on basic things on DSP
and learning concepts of FT (FFT and DFT), convolution etc...
At the same time I have also started get myself familiar with matlab's
signal processing toolbox.

I would appreciate if you can give sample programs in Matlab..
i) Generates the different signals
ii) Application of 3 parallel bandpass on each signal
iv) Application of some XYZ function on the frequency bins for removal
of undesired frequency removal

thanks in advance..

abdul n khan
a...@ieee.org

"Bevan Weiss" <k...@NOSPAMhotmail.com> wrote in message
news:<xf_0c.8020$S...@news.xtra.co.nz>...
> Generating a signal can be the easy part, depending on what kind of
> constraints you wish to place on it.
> 
> You could generate the data point values on the fly by using some kind of
> taylor series, or you could use a look up table with some kind of phase
> accumulator (my preferred option if you're generating multiple waveforms at
> different frequencies).  You could always use an analog sinusoidal
> oscillator and then some ADC to sample the waveform.  Any of these methods
> will generate a 'simple' sinusoid.
> 
> I think probably the best method would be for you to get a simple
> development board with a big player DSP on it, and several channels of ADC
> and DAC onboard.  You will also want some signal generators, perhaps you
> could make these yourself using some basic circuitry (assuming you're not
> doing anything high end just yet).
> You can then just play around with inputting several different frequency
> signals and trying to seperate them out into their own channels.  See what
> kind of effect the frequency spacing between the input signals is, and what
> effect it has on the seperated channels.
> It would give you a chance to experiment with the different filter methods,
> as well as other DSP topics.
> 
> 
> "Abdul Nasir Khan" <a...@ieee.org> wrote in message
> news:9...@posting.google.com...
> > But certainly, your mail is of very much help to me. I will now study
> > i) Band Pass Filter
> > ii) Application FIR, when signal is in time domain
> > iii) Multiplication of some XYZ function after applying FT on time
> > domain data.
> >
> > All of this expercise is only for my self-learning and is not
> > associated any real project at all. Therefore At the moment I myself
> > don't know to what kind of signals is it? In fact I will have to
> > generate my own signals, but how?.
> >
> > ...I am like working in the dark room and trying to look for a lost
> > needle.
> >
> > Any help will be appreciated in this context...
> >
> > Thanks
> >
> > Abdul N Khan
> > a.nasir @ ieee org
> >
> > "Bevan Weiss" <k...@NOSPAMhotmail.com> wrote in message
>  news:<3YO0c.7372$S...@news.xtra.co.nz>...
> > > You're not going to get a psuedo code algorithm for this...
> > > There are just too many ways of going about it.
> > >
> > > You are looking to essentially create three seperate parallel bandpass
> > > filters, each one will let through a single signal whilst blocking the
> > > remaining two signals.  Hence from the output of the three bandpass
>  filters
> > > you will have three seperated signals.
> > > We would need more information about the signals in particular to give
>  more
> > > information about whether it's an easy task to achieve this filtering.
> > > You could do this in the tiem domain by using a FIR or IIR filter, or
>  you
> > > could do it in the frequency domain by performing an FFT (or DFT) on the
> > > time domain data then multiplying the frequency bins by some function to
> > > remove the undesired frequencies, then simply perform an IFFT (or IDCT)
>  to
> > > get back into the time domain.
> > >
> > > If you perform a google search for this you should get plenty of links
>  to
> > > allow you to perform some research of your own.
> > >
> > > "Abdul Nasir Khan" <a...@ieee.org> wrote in message
> > > news:9...@posting.google.com...
> > > > Well..I am very new to DSP area. I would appreciate algorithm in the
> > > > form of psedue code.
> > > >
> > > > thanks and sincerely,
> > > >
> > > > abdul nasir khan
> > > >
> > > > "Mike Yarwood" <m...@btopenworld.com> wrote in message
>  news:<c1nkrg$5dp$1...@titan.btinternet.com>...
> > > > > If the three signals are in different frequency bands and those
>  bands
>  don't
> > > > > overlap then you should be able to separate them by filtering
>  without
> > > > > worrying too much about the amplitude - is this the case?
> > > > >
> > > > > "Abdul Nasir Khan" <a...@ieee.org> wrote in message
> > > > > news:9...@posting.google.com...
> > > > > > How can i separate three signals based on frequency and amplitude?
> > > > > >
> > > > > > Thanks
> > > > > >
> > > > > > abdul khan

a...@ieee.org (Abdul Nasir Khan) wrote in message
news:<9...@posting.google.com>...
> I am working at very basic level.
> My ultimate objective is implementing a project for learning that
> separates the noises from sound captured.
> 
> At the moment I am trying to get myself started on basic things on DSP
> and learning concepts of FT (FFT and DFT), convolution etc...
> At the same time I have also started get myself familiar with matlab's
> signal processing toolbox.
> 
> I would appreciate if you can give sample programs in Matlab..
> i) Generates the different signals

That's the easy part:

fs=1000;                        % sampling frequency 
N = 10000;                      % length of signal
tv= reshape([0:N-1]/fs,N,1);    % time line as column vector 

av=[1;2;3];                     % signal amplitudes as column vector
fv=[100,250,450];               % signal frequencies as row vector

s=sin(2*pi/fs*tv*fv)*a/sqrt(N); % generate signal

> ii) Application of 3 parallel bandpass on each signal

Check out a book on signal processing, like 

Lyons: "Understanding Digital Signal Processing"

You might find figure 6.38 in the 1st edition useful 
for learning about parallel filters. You might also 
want to look up "parallel filter bank" in other books.

> iv) Application of some XYZ function on the frequency bins for removal
> of undesired frequency removal

I don't understand what you mean here?

Rune