>i would say no..pure binary signal cannot be a cyclostationary
>signal..and regarding detecting a weak signal..i guess Cyclo spectral
>density (csd) and other cyclostationary related methods (cyclic
>autocorrelation) will be applied to find out weak signal in the
>noise..where (as per what you told) modulated signals will peak
>out..leaving out noise..
>
>

I've been reading up on cyclostationary signals recently and decided to
revive this thread. I was interested in the above-mentioned methods for
signal detection:
(a) "Cyclo Spectral Density"
(b) "Cyclic Autocorrelation"

For (a), is this the same as the "Spectral Correlation density" (SCD)? For
(b), what features can be seen from the cyclic autocorrelation that help
pull out the signal? (I assume similar to your ol' fashioned time-based
autocorrelation?  Is there a good reference to describe some of these
known-signal-extraction techniques (i.e. a Gardner reference of sorts)?

Last silly question. Can anyone provide a simple-intuitive explanation of
"cycle frequency"? The texts I found don't give much detail but rather just
SCDs of known plots with it depicted.

Thanks.

zsAJ1A  <a href="http://shcpuwtyzamf.com/">shcpuwtyzamf</a>, 
[url=http://ltxbkhlhmnxn.com/]ltxbkhlhmnxn[/url], 
[link=http://nhdgnmsshwse.com/]nhdgnmsshwse[/link], 
http://hdzvdpfwfxui.com/

>Hmmm.... this is new to me.  Can you share some references, please?

Julius, here is one:

Two alternative philosophies for estimation of the parameters of
time-series
Gardner, W.A.;
Information Theory, IEEE Transactions on
Volume 37,  Issue 1,  Jan. 1991 Page(s):216 - 218 

Enjoy,  :-)

Emre

On Nov 1, 3:02 pm, "emre" <egu...@ece.neu.edu> wrote:
>
> Hey Julius,
>
> Cyclostationarity does not imply randomness, unlike stationarity. (I just
> learned about this, too.  You can find this with a quick search on the
> web.)  Vladimir's statement only says "a binary signal".

Hmmm.... this is new to me.  Can you share some references, please?
As far as I know, a cyclostationary process is a stationary process
which statistical properties are invariant to time shifts by integer
multiples
of some value T.

So a cyclostationary process has to first be a stationary (wide-sense
or
strict sense) process, hence it has to be random in the first place.

>
> >I think that VLV means nT for n \in \Integers.
>
> The unit step function, u(t) above, satisfies this.  It changes value only
> for n = 0, which is an integer.
>
>  I just don't believe in the correctness of the statement. I am not trying
> to nit-pick.  :-)
>

Understood :-).  I'd love to see your reference on this, though.

Thanks,
Julius

On Nov 2, 1:02&#4294967295;am, "emre" <egu...@ece.neu.edu> wrote:
> >> >If a binary signal can change its value only at nT, then the signal is
> >> >cyclostationary.
>
> >> >VLV
>
> >> Wrong. &#4294967295;Counterexample: &#4294967295;unit step function, i.e., u(t) = 0 for t<0,
> and
> >> u(t)=1 for t>=0.
>
> >> Emre
>
> >But your example is not a stochastic process :-). &#4294967295;How can it be
> >stationary or cyclostationary or non-stationary?
>
> Hey Julius,
>
> Cyclostationarity does not imply randomness, unlike stationarity. (I just
> learned about this, too. &#4294967295;You can find this with a quick search on the
> web.) &#4294967295;Vladimir's statement only says "a binary signal".
>
> >I think that VLV means nT for n \in \Integers.
>
> The unit step function, u(t) above, satisfies this. &#4294967295;It changes value only
> for n = 0, which is an integer.
>
> &#4294967295;I just don't believe in the correctness of the statement. I am not trying
> to nit-pick. &#4294967295;:-)
>
> Emre

plz refer to these files on matlab implementationand you can quickly
check the binary signal also.

http://www.mathworks.fr/matlabcentral/fileexchange/16742

>> >If a binary signal can change its value only at nT, then the signal is
>> >cyclostationary.
>>
>> >VLV
>>
>> Wrong.  Counterexample:  unit step function, i.e., u(t) = 0 for t<0,
and
>> u(t)=1 for t>=0.
>>
>> Emre
>
>But your example is not a stochastic process :-).  How can it be
>stationary or cyclostationary or non-stationary?

Hey Julius,

Cyclostationarity does not imply randomness, unlike stationarity. (I just
learned about this, too.  You can find this with a quick search on the
web.)  Vladimir's statement only says "a binary signal".

>I think that VLV means nT for n \in \Integers.

The unit step function, u(t) above, satisfies this.  It changes value only
for n = 0, which is an integer.

 I just don't believe in the correctness of the statement. I am not trying
to nit-pick.  :-)

Emre

On Nov 1, 10:50 am, "emre" <egu...@ece.neu.edu> wrote:
> >If a binary signal can change its value only at nT, then the signal is
> >cyclostationary.
>
> >VLV
>
> Wrong.  Counterexample:  unit step function, i.e., u(t) = 0 for t<0, and
> u(t)=1 for t>=0.
>
> Emre

But your example is not a stochastic process :-).  How can it be
stationary or cyclostationary or non-stationary?

I think that VLV means nT for n \in \Integers.

To the original poster, this is a well-answered question in just about
any digital communication book.

>If a binary signal can change its value only at nT, then the signal is 
>cyclostationary.
>
>VLV

Wrong.  Counterexample:  unit step function, i.e., u(t) = 0 for t<0, and
u(t)=1 for t>=0.

Emre

>>
>>
>>PARTICLEREDDY (STRAYDOG) wrote:
>>
>>
>>> i would say no..pure binary signal cannot be a cyclostationary
>>> signal..
>>
>>????
>>
>>If a binary signal can change its value only at nT, then the signal is 
>>cyclostationary.
>>
>>VLV
>>
>>
>
>Mmmm.... I see your point here. I would agree with you based on mu basic
>understanding to cyclostationary (process with periodic statistics) but
how
>can you prove that? any reference to understand that better? Why digital
>communication text boox assume a normally distributed data source, or
even
>more than that, the basic approximation of the quantization error is a
>normal distributed noise (found to be not accurate but still alot of
people
>accept that). 
>
>The point about detection using scd is absolutely a good point since I
>have read that for signal detection. 
>
>More clarification will be appreciated VLV :)
>
>

Let me also elaborate on my comment on modeling the quantization error as
normally distributed noise (Stationary process.) The quantization noise can
be seen as a binary signal as well (in away except that it has multi-level
rather than binary level but I do not feel that the level issue will affect
the modeling as a stationary or cyclostationary signal). If that is true,
then this will be the reason I thought that binary signals are stationary
process. 

VLV, I am very interested to know from you your clarification on this
point and thanks in advance sharing this knowedge with the group.

>
>
>PARTICLEREDDY (STRAYDOG) wrote:
>
>
>> i would say no..pure binary signal cannot be a cyclostationary
>> signal..
>
>????
>
>If a binary signal can change its value only at nT, then the signal is 
>cyclostationary.
>
>VLV
>
>

Mmmm.... I see your point here. I would agree with you based on mu basic
understanding to cyclostationary (process with periodic statistics) but how
can you prove that? any reference to understand that better? Why digital
communication text boox assume a normally distributed data source, or even
more than that, the basic approximation of the quantization error is a
normal distributed noise (found to be not accurate but still alot of people
accept that). 

The point about detection using scd is absolutely a good point since I
have read that for signal detection. 

More clarification will be appreciated VLV :)