AWR6843AOP: The cloud point of child will be disappeared after another adult enter the car

Hey Chou,

Thanks for reaching out! One of our experts on child presence detection will take a look into this within the next day or two.

Regards,

Kristien

HI, Chou:

Can you let me know which TI demo you are using for this experiment?  

Best,

Zigang

Hi Zigang,

Very thanks for your help.

We use the lab example of InCabin_Sesnsing(3D_Occupancy_Detection_Classification) as the following screenshot.

And, the child is using Ashton-Drake "Bella Rose Baby Doll". It's the same with TI's testing baby doll.
( https://www.ashtondrake.com/products/301881001_lifelike-breathing-baby-doll.html )

Sincerely,

TC.Chou

HI, Chou:

It is normal that in the range-angle domain, a strong reflection can shadow a weak reflection if they are in the similar range.  It may not be a CFAR issue, because the heatmap itself may not show a peak from the weak target. 

The only thing I can suggest is to try is try to adjust the parameter related to the angle domain.  Below highlighted all the average window size, guard window size, detection threshold and sidelobe threshold, that are related to the angle domain.  And you can try to play with it.  You can modify the K0 and sidelobe threshold first.   The definition of each parameter can be found at: AWR6843 CPD With Classification Users Guide

 [Original] dynamicRACfarCfg -1 5 29 1 1 8 "8" 4  6 "4" 1 8.00 "6.00" "0.40" 1 15
 [New]      dynamicRACfarCfg -1 5 29 1 1 8 "4" 4  6 "2" 1 8.00 "5.00" "0.30" 1 15

But as I said, the heatmap may not show a peak from the weaker reflection when there is a strong reflection is present in the same range bin.

Best,

Zigang

Hi Zigang,

Many thanks for your help.

It seems no significant improvement after we use your suggested dynamicRACfarCfg -1 5 29 1 1 8 "4" 4  6 "2" 1 8.00 "5.00" "0.30" 1 15.

Current, we record Mathlab GUI Video (below) and do some analysis: (https://www.asuswebstorage.com/navigate/a/#/s/2B74CBE901044932858EF30752BD54A3Y)

- [Our Scenario Description]:

Step 1). At first, a "Bella Rose Baby Doll" is put on Seat No.5.

Step 2). Then, an adult sits on Seat No.3.

Step 3). Finally, an adult leaves Seat No.3.

- [Our Analysis]

(a). You can see if the [cloud point and SNR] of adult is relatively "LOW", both an adult cloud point and a baby doll cloud point will appear on the same time.

(b). You can see if the [cloud point and SNR] of adult is relatively "HIGH", only an adult cloud point will appear, and a baby doll cloud point disappear.

Hence, we think it may be caused by the average of SNR(reference windows) is larger than target point of baby doll when we run CASO-CFAR algorithm.  If our viewpoint is correct, is it possible to set the max limitation of the average of SNR(reference windows) to avoid filter baby doll.

In chirp configuration of " occStateMach  0 10 10 3 50 1 3 12 20 4 700.0".

--threshold1[10 10] => 10 point and average SNR > 10

--threshold2[3 50] => 3 point  and average SNR > 50

Therefore, can we setup the max limitation of the average of SNR(reference windows)  based on threshold1 and threshold2.

Sincerely,

TC.Chou

HI, Chou:

The occStateMach is the state machine threshold after point cloud.  It contains the threshold for detecting the occupancy based on point cloud.  IF we see some points clouds but did not see the occupancy, then you can consider tuning the occStateMach.  But right now, your problem is that you are not seeing point cloud for the baby doll.  In this case, tuning occStateMach will not be helpful.

You can reduce the sidelobe threshold further to give a try.  But as I said, it may not be helpful since the strong signal can mask the weak signal and there may not be any peak from baby showing in the range-angle heatmap.   

dynamicRACfarCfg -1 5 29 1 1 8 4 4  6 2 1 8.00 "5.00" "0.10" 1 15

Best,

Zigang