Forums

Energy detector

Started by Krishna_11_05 April 26, 2008
We do have a 32-bit floating point processor. ADSP 21160. I have used
energy detection because I thought it was the best for detecting unknown
signals as opposed to cross-correlation in active mode. So I may be wrong.
Is it so?
On Apr 26, 4:12 am, "Krishna_11_05" <mv_mail_box-jkri...@yahoo.co.in>
wrote:
> We do have a 32-bit floating point processor. ADSP 21160. I have used > energy detection because I thought it was the best for detecting unknown > signals as opposed to cross-correlation in active mode. So I may be wrong. > Is it so?
So, what do you know and not know about the 'unknown signal'? Time of arrival, duration, frequency, bandwidth, amplitude, modulation? How many signals can arrive at once? How many different types of signals do you want to use your detector for? What are the number of channels and bandwidths? What are the computational resources available? To what do we relate the meaning of 'best'? What did you think of Nuttall's power-law detector approach when you Googled on it and downloaded the free documentation on it? Does it seem like what you are thinking of doing? Why did you start a new thread instead of continuing the previous one? Dale B. Dalrymple
On Apr 26, 3:03 pm, dbd <d...@ieee.org> wrote:
> On Apr 26, 4:12 am, "Krishna_11_05" <mv_mail_box-jkri...@yahoo.co.in> > wrote: > > > We do have a 32-bit floating point processor. ADSP 21160. I have used > > energy detection because I thought it was the best for detecting unknown > > signals as opposed to cross-correlation in active mode. So I may be wrong. > > Is it so? > > So, what do you know and not know about the 'unknown signal'? Time of > arrival, duration, frequency, bandwidth, amplitude, modulation? How > many signals can arrive at once? How many different types of signals > do you want to use your detector for? What are the number of channels > and bandwidths? What are the computational resources available? To > what do we relate the meaning of 'best'? > > What did you think of Nuttall's power-law detector approach when you > Googled on it and downloaded the free documentation on it? Does it > seem like what you are thinking of doing? > > Why did you start a new thread instead of continuing the previous one? > > Dale B. Dalrymple
I'm interested in this topic. I did google it but didn't find much for free, at least not in the first few pages. I'll check IEEE on Monday when I get back to work. John
On Apr 26, 5:52 pm, John <sampson...@gmail.com> wrote:
> On Apr 26, 3:03 pm, dbd <d...@ieee.org> wrote: > > > > > On Apr 26, 4:12 am, "Krishna_11_05" <mv_mail_box-jkri...@yahoo.co.in> > > wrote: > > > > We do have a 32-bit floating point processor. ADSP 21160. I have used > > > energy detection because I thought it was the best for detecting unknown > > > signals as opposed to cross-correlation in active mode. So I may be wrong. > > > Is it so? > > > So, what do you know and not know about the 'unknown signal'? Time of > > arrival, duration, frequency, bandwidth, amplitude, modulation? How > > many signals can arrive at once? How many different types of signals > > do you want to use your detector for? What are the number of channels > > and bandwidths? What are the computational resources available? To > > what do we relate the meaning of 'best'? > > > What did you think of Nuttall's power-law detector approach when you > > Googled on it and downloaded the free documentation on it? Does it > > seem like what you are thinking of doing? > > > Why did you start a new thread instead of continuing the previous one? > > > Dale B. Dalrymple > > I'm interested in this topic. I did google it but didn't find much for > free, at least not in the first few pages. I'll check IEEE on Monday > when I get back to work. > > John
For Nuttall power-law detector I find 3 of the first 5 are freely downloadable. One from citeceer, two from stinet. There's another one from DSTO within the first ten. YMMV. The two on stinet are Nuttall's original reports. These are also available from ds.scientificcommons.org further on. Dale B. Dalrymple
>On Apr 26, 4:12 am, "Krishna_11_05" <mv_mail_box-jkri...@yahoo.co.in> >wrote: >> We do have a 32-bit floating point processor. ADSP 21160. I have used >> energy detection because I thought it was the best for detecting
unknown
>> signals as opposed to cross-correlation in active mode. So I may be
wrong.
>> Is it so? > >So, what do you know and not know about the 'unknown signal'? Time of >arrival, duration, frequency, bandwidth, amplitude, modulation? How >many signals can arrive at once? How many different types of signals >do you want to use your detector for? What are the number of channels >and bandwidths? What are the computational resources available? To >what do we relate the meaning of 'best'? > >What did you think of Nuttall's power-law detector approach when you >Googled on it and downloaded the free documentation on it? Does it >seem like what you are thinking of doing? > >Why did you start a new thread instead of continuing the previous one? > >Dale B. Dalrymple >
Thanks for the reply. I am sorry about the new thread. I am not a regular here and so I goofed. The unknown signal is basically unknown. I mean it can have any duration, any repetition interval, any amplitude. I only know that it can come in bursts and is not a continuous signal. I am trying to use this detector for only two types of signals but how does that matter. In energy detector, I can get the burst only. After that I analyze the burst for any interesting information. There is only a single channel. Omni. Computational resources are not a limitation. I did not get any free Nuttall documenation. I will try the links you suggested. Thanks. Krishna.
On Apr 27, 5:39 am, "Krishna_11_05" <mv_mail_box-jkri...@yahoo.co.in>
wrote:
> >On Apr 26, 4:12 am, "Krishna_11_05" <mv_mail_box-jkri...@yahoo.co.in> > >wrote: > >> We do have a 32-bit floating point processor. ADSP 21160. I have used > >> energy detection because I thought it was the best for detecting > unknown > >> signals as opposed to cross-correlation in active mode. So I may be > wrong. > >> Is it so? > > >So, what do you know and not know about the 'unknown signal'? Time of > >arrival, duration, frequency, bandwidth, amplitude, modulation? How > >many signals can arrive at once? How many different types of signals > >do you want to use your detector for? What are the number of channels > >and bandwidths? What are the computational resources available? To > >what do we relate the meaning of 'best'? > > >What did you think of Nuttall's power-law detector approach when you > >Googled on it and downloaded the free documentation on it? Does it > >seem like what you are thinking of doing? > > >Why did you start a new thread instead of continuing the previous one? > > >Dale B. Dalrymple > > Thanks for the reply. I am sorry about the new thread. I am not a regular > here and so I goofed. > The unknown signal is basically unknown. I mean it can have any duration, > any repetition interval, any amplitude. I only know that it can come in > bursts and is not a continuous signal. I am trying to use this detector for > only two types of signals but how does that matter. In energy detector, I > can get the burst only. After that I analyze the burst for any interesting > information. There is only a single channel. Omni. Computational resources > are not a limitation. > I did not get any free Nuttall documenation. I will try the links you > suggested. Thanks. > Krishna.
Krishna Do you run your detector on time series data? Do you perform an FFT on it first? If you do an FFT, how big? And how did you decide to use that size? How many bins does the signal appear in? These influence what detector structure you could use. John The English language version should be: http://en.scientificcommons.org They seem to have picked up some US and Australian literature but not the Canadian yet. Dale B. Dalrymple
On Apr 27, 12:20 am, dbd <d...@ieee.org> wrote:
> On Apr 26, 5:52 pm, John <sampson...@gmail.com> wrote: > > > > > On Apr 26, 3:03 pm, dbd <d...@ieee.org> wrote: > > > > On Apr 26, 4:12 am, "Krishna_11_05" <mv_mail_box-jkri...@yahoo.co.in> > > > wrote: > > > > > We do have a 32-bit floating point processor. ADSP 21160. I have used > > > > energy detection because I thought it was the best for detecting unknown > > > > signals as opposed to cross-correlation in active mode. So I may be wrong. > > > > Is it so? > > > > So, what do you know and not know about the 'unknown signal'? Time of > > > arrival, duration, frequency, bandwidth, amplitude, modulation? How > > > many signals can arrive at once? How many different types of signals > > > do you want to use your detector for? What are the number of channels > > > and bandwidths? What are the computational resources available? To > > > what do we relate the meaning of 'best'? > > > > What did you think of Nuttall's power-law detector approach when you > > > Googled on it and downloaded the free documentation on it? Does it > > > seem like what you are thinking of doing? > > > > Why did you start a new thread instead of continuing the previous one? > > > > Dale B. Dalrymple > > > I'm interested in this topic. I did google it but didn't find much for > > free, at least not in the first few pages. I'll check IEEE on Monday > > when I get back to work. > > > John > > For > > Nuttall power-law detector > > I find 3 of the first 5 are freely downloadable. One from citeceer, > two from stinet. There's another one from DSTO within the first ten. > YMMV. > > The two on stinet are Nuttall's original reports. These are also > available from ds.scientificcommons.org further on. > > Dale B. Dalrymple
Aha, I found them, thank you! John
>>Do you run your detector on time series data? Do you perform an FFT on >it first? > >If you do an FFT, how big? And how did you decide to use that size? >How many bins does the signal appear in? These influence what detector >structure you could use. > > >John > >The English language version should be: > >http://en.scientificcommons.org > >They seem to have picked up some US and Australian literature but not >the Canadian yet. > >Dale B. Dalrymple >
Yes. The detector is run on a time series data. No FFT is done. The raw data after dynamic range compression is digitized and energy is computed. It is compared to threshold which is also normalized energy of a large data (3 times the amximum expected burs duration). I am taking energy for very small time so that I get the time of arrival accurately. This does reduce the SNR, but may be I have it. Atleast during testing.