• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 30 subscribers
  • Views 629 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

IWR1642BOOST: People Counting Lab - Allocation Parameters

Boaz Menis
Intellectual 510 points
Part Number: IWR1642BOOST

Hi,

In the people counting lab there is set of parameters called "Allocation Parameters".

one of the parameters in this set is called "SNR threshold".

If I'm not mistaking, this is the minimum linear sum of the cluster SNR in order it to be allocated to some target.

The thing is, that the SNR is decreasing with the range of the target (SNR ~ Range^(-4)).

That means that if the range is doubled, the SNR decreases by a factor of 16 (=2^4).

I think that the right thing was to work with some kind of "RCS" threshold parameter instead of SNR parameter,

or some kind of other normalized SNR parameter.

Is the SNR actually normalized by the range ?  I just didn't see it in the code...

Thanks,

Boaz

 

  • 0
    TI__Mastermind 28055 points
    Hi Boaz,

    The SNR value for each point is not true SNR. Instead, we divide the signal intensity at the peak (detected point) by the average value of the bins around it - this is done in CFAR detection. This "SNR" is more of a power ratio, and drops off much more slowly with range.

    Regards,
    Justin
  • 0 in reply to Justin Curewitz
    Intellectual 510 points
    Hi Justin,

    I understand. But still, the CFAR estimates the noise level in a certain bin range so taking he ratio between the peak value (squared) to
    the noise level is your best estimation of SNR.
    I did some experiments. For example, when I sit behind the radar (so it can't detect me) and I just wave my hand in front of the radar
    it will count me as a person, but, when I stand (and move) in front of the radar in a distance of 10-14 meters (with a proper configuration)
    the radar detects me but it won't allocate my cluster since it won't reach the SNR threshold (the default parameter = 100).
    The moment I decreased the threshold to 0 (but keeping the minimum number of points for cluster to be around 5) it worked.
    I believe there is some normalization with range that should be done in order to keep the algorithm consistent for various ranges.

    Regards,
    Boaz
  • 0 in reply to Boaz Menis
    TI__Mastermind 28055 points
    Hi Boaz,

    If you are interested in modifying the demo to incorporate this feature, you can find the code at:
    <MMWAVE_Industrial_Toolbox_VERSION>\labs\lab0011-pplcount\lab0011_pplcount_pjt\radarDemo\modules\detection\CFAR\src
    The CFAR calls which return the "SNR" are defined here.

    Do you have any more questions regarding the demo?

    Regards,
    Justin