Started by September 3, 2008
```hi
i am working on the processing of radar signals,, i have a constant
PRI(pulse repitition interval) radar and i take (say 40)pulses , calculate
their difference histogram to identify peaks(in a histogram), from peak i
can calculate the incoming pulses' PRI

pulses, its difference histogram is spread-out, and the peaks are not
obvious,

how can i guess the PRI from this histogram or is there any method of
correcting the histogram spread-out due to noise ??

umair

```
```On Sep 3, 8:10 am, "umair039" <umair...@yahoo.com> wrote:
> hi
> i am working on the processing of radar signals,, i have a constant
> PRI(pulse repitition interval) radar and i take (say 40)pulses , calculate
> their difference histogram to identify peaks(in a histogram), from peak i
> can calculate the incoming pulses' PRI
>
> pulses, its difference histogram is spread-out, and the peaks are not
> obvious,
>
> how can i guess the PRI from this histogram or is there any method of
> correcting the histogram spread-out due to noise ??
>
> umair

First, are you using a matched filter or a moving average filter to
increase your SNR? Also, how are you calculating the pulse threshold?
Do you have to recognize different PRIs simultaneously (due to

Darol Klawetter
```
```>On Sep 3, 8:10 am, "umair039" <umair...@yahoo.com> wrote:
>> hi
>> i am working on the processing of radar signals,, i have a constant
>> PRI(pulse repitition interval) radar and i take (say 40)pulses ,
calculate
>> their difference histogram to identify peaks(in a histogram), from peak
i
>> can calculate the incoming pulses' PRI
>>
>> pulses, its difference histogram is spread-out, and the peaks are not
>> obvious,
>>
>> how can i guess the PRI from this histogram or is there any method of
>> correcting the histogram spread-out due to noise ??
>>
>> umair
>
>
>First, are you using a matched filter or a moving average filter to
>increase your SNR? Also, how are you calculating the pulse threshold?
>Do you have to recognize different PRIs simultaneously (due to
>
>Darol Klawetter

dear darol
i am not using any filter to increase the SNR, my task is to only detect
the PRI whose histogram is spread-out after addition of (10% of PRI)
gaussian noise,  i have to just make some intelligent observations through
the spread-out histogram as what portion of histogram really is indicative
of the incoming noisy PRI,(in case of original PRI histogram with no noise
the peak's index and envelope in the histogram signifies the desired).

the pulse threshold is calculated by the formula:
needed_pulses =  THRESHOLD * (win_size/(max_index*HistogramResolution));
where , needed_pulses = minimum 70% of pulses
THRESHOLD = 0.7
win_size = observation window which has 40 pulses of incoming PRI
max_index = peak index in a histogram
HistogramResolution = 20;

currently i am working on the detection of only single PRI(in noisy
analysis) , later i will improve my algo to include more PRIs and process
those with noise aswell.
refer to US Patent 5583505 for details, (www.patentstorm.us)

thanks again and takecare
umair
```
```On 3 Sep, 18:07, Darol Klawetter <darol.klawet...@l-3com.com> wrote:
> On Sep 3, 8:10 am, "umair039" <umair...@yahoo.com> wrote:
>
> > hi
> > i am working on the processing of radar signals,, i have a constant
> > PRI(pulse repitition interval) radar and i take (say 40)pulses , calculate
> > their difference histogram to identify peaks(in a histogram), from peak i
> > can calculate the incoming pulses' PRI
>
> > i want to ask if i add a gaussian noise in my incoming radar signal
> > pulses, its difference histogram is spread-out, and the peaks are not
> > obvious,
>
> > how can i guess the PRI from this histogram or is there any method of
> > correcting the histogram spread-out due to noise ??
>
> > umair
>
> First, are you using a matched filter or a moving average filter to
> increase your SNR? Also, how are you calculating the pulse threshold?
> Do you have to recognize different PRIs simultaneously (due to
> multiple radars on same frequency)?
>
> Darol Klawetter

Sounds like he wants to detect all radars on same frequency, where
individual useful radars. Maybe the return ping is less useful at
locating than the out ping in such a transport lane due to r2 versus
r4.

cheers
jacko

p.s. what do you think he wants?
```
```On Sep 3, 10:10 pm, "umair039" <umair...@yahoo.com> wrote:
> >On Sep 3, 8:10 am, "umair039" <umair...@yahoo.com> wrote:
> >> hi
> >> i am working on the processing of radar signals,, i have a constant
> >> PRI(pulse repitition interval) radar and i take (say 40)pulses ,
> calculate
> >> their difference histogram to identify peaks(in a histogram), from peak
> i
> >> can calculate the incoming pulses' PRI
>
> >> i want to ask if i add a gaussian noise in my incoming radar signal
> >> pulses, its difference histogram is spread-out, and the peaks are not
> >> obvious,
>
> >> how can i guess the PRI from this histogram or is there any method of
> >> correcting the histogram spread-out due to noise ??
>
> >> umair
>
> >First, are you using a matched filter or a moving average filter to
> >increase your SNR? Also, how are you calculating the pulse threshold?
> >Do you have to recognize different PRIs simultaneously (due to
> >multiple radars on same frequency)?
>
> >Darol Klawetter
>
> dear darol
> i am not using any filter to increase the SNR, my task is to only detect
> the PRI whose histogram is spread-out after addition of (10% of PRI)
> gaussian noise,  i have to just make some intelligent observations through
> the spread-out histogram as what portion of histogram really is indicative
> of the incoming noisy PRI,(in case of original PRI histogram with no noise
> the peak's index and envelope in the histogram signifies the desired).
>
> the pulse threshold is calculated by the formula:
> needed_pulses =  THRESHOLD * (win_size/(max_index*HistogramResolution));
> where , needed_pulses = minimum 70% of pulses
> THRESHOLD = 0.7
> win_size = observation window which has 40 pulses of incoming PRI
> max_index = peak index in a histogram
> HistogramResolution = 20;
>
> currently i am working on the detection of only single PRI(in noisy
> analysis) , later i will improve my algo to include more PRIs and process
> those with noise aswell.
> refer to US Patent 5583505 for details, (www.patentstorm.us)
>
> thanks again and takecare
> waiting for ur further reply
> umair

It appears that you're adding a jitter of 10% of the nominal PRI, and
the distribution of the jitter within this 10% bounds is gaussian,
correct? If so, it seems that you could, with a large enough sampling
of pulses, see a gaussian distribution in your histogram and thus you
could identify the peak, and thus nominal PRI. What I'm I missing?

Darol Klawetter
```