Prev: Generate Alternate +1 & -1 sequence
Next: GA Tool to solve TSP
From: AB on 29 Mar 2010 07:30
Hi everybody,
I would like to generate two sets of complex value signal which is
the first set is totally uncorelated real and imaginary parts and the second set
is in opposite totally correlate real and imaginary parts.

Let say the signals X1= is of size (2,2000). I used the rand cmd to generate two real signals Xr and Xi and combine both for the first set X1=Xr+i*Xi.
However I'm not really sure for the second set since I need to make it both correlate in terms of amplitude and phase as well. Anybody outside there can help me?

Every comment and suggestions will be very helpful.
AB


From: Wayne King on 29 Mar 2010 08:24
"AB " <amran_abdhadi(a)yahoo.com> wrote in message <hoq30c$1i1$1(a)fred.mathworks.com>...
> Hi everybody,
> I would like to generate two sets of complex value signal which is
> the first set is totally uncorelated real and imaginary parts and the second set
> is in opposite totally correlate real and imaginary parts.
>
> Let say the signals X1= is of size (2,2000). I used the rand cmd to generate two real signals Xr and Xi and combine both for the first set X1=Xr+i*Xi.
> However I'm not really sure for the second set since I need to make it both correlate in terms of amplitude and phase as well. Anybody outside there can help me?
>
> Every comment and suggestions will be very helpful.
> AB
>
>

Hi,
One option:
if you have the Statistics Toolbox, you can use mvnrnd to generate vectors from a multivariate normal with a specified covariance matrix and then make one the real and one the imaginary part.

Another option is filtering. If you filter white noise, you will "induce" correlation.

Not sure from your post if you have some specific form of cross correlation between the real and imaginary parts you are trying to produce, or just that the cross correlation should be significantly different from zero.

Wayne
From: AB on 29 Mar 2010 11:20
Hi Wayne,

Thank you for your suggestion. I've tried the mvnrnd to generate the signal.However it just allow me to use sigma(cov) as semi-definite matrix.

What I really need is to variate the correlation between the imag and real part of the complex signal. For your info, in my area of study , most method consider the real and imag part are uncorrelated (or they called it circular source). However I want to investigate the effect of the correlation to the algorithm that I'm developed. That's why I need to create the mentioned signals with certain level of correlation.

thank you.
AB

"Wayne King" <wmkingty(a)gmail.com> wrote in message <hoq657$krc$1(a)fred.mathworks.com>...
> "AB " <amran_abdhadi(a)yahoo.com> wrote in message <hoq30c$1i1$1(a)fred.mathworks.com>...
> > Hi everybody,
> > I would like to generate two sets of complex value signal which is
> > the first set is totally uncorelated real and imaginary parts and the second set
> > is in opposite totally correlate real and imaginary parts.
> >
> > Let say the signals X1= is of size (2,2000). I used the rand cmd to generate two real signals Xr and Xi and combine both for the first set X1=Xr+i*Xi.
> > However I'm not really sure for the second set since I need to make it both correlate in terms of amplitude and phase as well. Anybody outside there can help me?
> >
> > Every comment and suggestions will be very helpful.
> > AB
> >
> >
>
> Hi,
> One option:
> if you have the Statistics Toolbox, you can use mvnrnd to generate vectors from a multivariate normal with a specified covariance matrix and then make one the real and one the imaginary part.
>
> Another option is filtering. If you filter white noise, you will "induce" correlation.
>
> Not sure from your post if you have some specific form of cross correlation between the real and imaginary parts you are trying to produce, or just that the cross correlation should be significantly different from zero.
>
> Wayne
From: Wayne King on 29 Mar 2010 12:44
"AB " <amran_abdhadi(a)yahoo.com> wrote in message <hoqgfl$nfq$1(a)fred.mathworks.com>...
> Hi Wayne,
>
> Thank you for your suggestion. I've tried the mvnrnd to generate the signal.However it just allow me to use sigma(cov) as semi-definite matrix.
>
> What I really need is to variate the correlation between the imag and real part of the complex signal. For your info, in my area of study , most method consider the real and imag part are uncorrelated (or they called it circular source). However I want to investigate the effect of the correlation to the algorithm that I'm developed. That's why I need to create the mentioned signals with certain level of correlation.
>
> thank you.
> AB
>
> "Wayne King" <wmkingty(a)gmail.com> wrote in message <hoq657$krc$1(a)fred.mathworks.com>...
> > "AB " <amran_abdhadi(a)yahoo.com> wrote in message <hoq30c$1i1$1(a)fred.mathworks.com>...
> > > Hi everybody,
> > > I would like to generate two sets of complex value signal which is
> > > the first set is totally uncorelated real and imaginary parts and the second set
> > > is in opposite totally correlate real and imaginary parts.
> > >
> > > Let say the signals X1= is of size (2,2000). I used the rand cmd to generate two real signals Xr and Xi and combine both for the first set X1=Xr+i*Xi.
> > > However I'm not really sure for the second set since I need to make it both correlate in terms of amplitude and phase as well. Anybody outside there can help me?
> > >
> > > Every comment and suggestions will be very helpful.
> > > AB
> > >
> > >
> >
> > Hi,
> > One option:
> > if you have the Statistics Toolbox, you can use mvnrnd to generate vectors from a multivariate normal with a specified covariance matrix and then make one the real and one the imaginary part.
> >
> > Another option is filtering. If you filter white noise, you will "induce" correlation.
> >
> > Not sure from your post if you have some specific form of cross correlation between the real and imaginary parts you are trying to produce, or just that the cross correlation should be significantly different from zero.
> >
> > Wayne

Hi AB, this does generate a complex-valued signal with correlated real and imaginary parts

R = mvnrnd([0 0],[1 .4; .4 1],100);
Z = R(:,1)+1j*R(:,2);

The other option I mentioned works too, you can simply use a filter.

Wayne
From: AB on 30 Mar 2010 04:54
Hi Wayne,
Thanks. I got it!

AB

"Wayne King" <wmkingty(a)gmail.com> wrote in message <hoqld7$hb6$1(a)fred.mathworks.com>...
> "AB " <amran_abdhadi(a)yahoo.com> wrote in message <hoqgfl$nfq$1(a)fred.mathworks.com>...
> > Hi Wayne,
> >
> > Thank you for your suggestion. I've tried the mvnrnd to generate the signal.However it just allow me to use sigma(cov) as semi-definite matrix.
> >
> > What I really need is to variate the correlation between the imag and real part of the complex signal. For your info, in my area of study , most method consider the real and imag part are uncorrelated (or they called it circular source). However I want to investigate the effect of the correlation to the algorithm that I'm developed. That's why I need to create the mentioned signals with certain level of correlation.
> >
> > thank you.
> > AB
> >
> > "Wayne King" <wmkingty(a)gmail.com> wrote in message <hoq657$krc$1(a)fred.mathworks.com>...
> > > "AB " <amran_abdhadi(a)yahoo.com> wrote in message <hoq30c$1i1$1(a)fred.mathworks.com>...
> > > > Hi everybody,
> > > > I would like to generate two sets of complex value signal which is
> > > > the first set is totally uncorelated real and imaginary parts and the second set
> > > > is in opposite totally correlate real and imaginary parts.
> > > >
> > > > Let say the signals X1= is of size (2,2000). I used the rand cmd to generate two real signals Xr and Xi and combine both for the first set X1=Xr+i*Xi.
> > > > However I'm not really sure for the second set since I need to make it both correlate in terms of amplitude and phase as well. Anybody outside there can help me?
> > > >
> > > > Every comment and suggestions will be very helpful.
> > > > AB
> > > >
> > > >
> > >
> > > Hi,
> > > One option:
> > > if you have the Statistics Toolbox, you can use mvnrnd to generate vectors from a multivariate normal with a specified covariance matrix and then make one the real and one the imaginary part.
> > >
> > > Another option is filtering. If you filter white noise, you will "induce" correlation.
> > >
> > > Not sure from your post if you have some specific form of cross correlation between the real and imaginary parts you are trying to produce, or just that the cross correlation should be significantly different from zero.
> > >
> > > Wayne
>
> Hi AB, this does generate a complex-valued signal with correlated real and imaginary parts
>
> R = mvnrnd([0 0],[1 .4; .4 1],100);
> Z = R(:,1)+1j*R(:,2);
>
> The other option I mentioned works too, you can simply use a filter.
>
> Wayne
 | 
Pages: 1
Prev: Generate Alternate +1 & -1 sequence
Next: GA Tool to solve TSP