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.

IWR1443BOOST: CFAR detection

Part Number: IWR1443BOOST
Other Parts Discussed in Thread: IWR6843, IWR1443

Hi,

1. Is CFAR applied to each doppler bin independently, where the peaks detected in each doppler bin are added to the detection output list? 

2. If the number of detections are over the maximum limit, how are the detections selected for pointcloud output? In my setup, the sensor is used to monitor persons in a room. If there are two persons walking in the room, say person1 and person2, person1 could have a large number of points detected on him and person2 very few points. When this happens, the number of points in the pointcloud are usually at the observed maximum limit (64). My guess is that when the number of detections have reached the maximum limit, the CFAR process stops, so part of person2's points are not extracted. Am I right on this? If so, can you give suggestions on how to enable a more balanced point spread between persons in the room?

Thank you.

  • Hi,

    Which lab are you using? It sounds like you're using the out-of-box demo, but the people counting or area scanner labs sound like better options here. These will likely require you to use an IWR6843 device though, which has a dedicated DSP. You should be able to detect and track more than enough people in the room with this. We are able to detect 10 with no problem.

    Best,

    Nate

  • I am running the out-of-box demo firmware on the IWR1443. I am using a software running on PC, and not demo visualizer, to process the data stream from the sensor.

    Both people counting and area scanner labs are not supported on IWR1443. In any case, the point cloud data is sufficient, we can handle the detection and tracking on the PC.

  • I would like to evaluate the feasibility of using IWR1443 first before considering other devices.

  • In that case, 

    1. Is CFAR applied to each doppler bin independently, where the peaks detected in each doppler bin are added to the detection output list? 

    What do you mean by this? If you want to learn about the implementation of the CFAR algorithm, consider reviewing 

    C:/ti/mmwave_sdk_03_05_00_04/packages/ti/datapath/dpc/dpu/cfarcaproc/docs/doxygen/html/index.html

    which discusses how the Doppler CFAR runs on the HWA and the parameter set that governs it.

    2. If the number of detections are over the maximum limit, how are the detections selected for pointcloud output? In my setup, the sensor is used to monitor persons in a room. If there are two persons walking in the room, say person1 and person2, person1 could have a large number of points detected on him and person2 very few points. When this happens, the number of points in the pointcloud are usually at the observed maximum limit (64). My guess is that when the number of detections have reached the maximum limit, the CFAR process stops, so part of person2's points are not extracted. Am I right on this? If so, can you give suggestions on how to enable a more balanced point spread between persons in the room?

    In principle, I understand your logic, but I think you can increase the maximum number of points beyond 64. Where did you get that maximum of 64 from?

    Best,

    Nate

  • 1. From the description of cfarCfg in SDK2.1 user guide, only range direction is supported for my device.  Which means CFAR is only done in range domain? Thus, for each dopplerIdx in the range-doppler heatmap, there would be zero of more CFAR detections. And all these detections would be added to the detection output list. Since my device is on SDK 2.1, I am inferring from the SDK2.1 demo document, C:/ti/mmwave_sdk_02_01_00_04/packages/ti/demo/xwr14xx/mmw/docs/doxygen/html/index.html. 

    2. The number 64 is based on the pointcloud output I obtained using various configurations. I am not able to increase the number of points above 64.

  • 1. Your analysis is correct.

    2. The number 64 is based on the pointcloud output I obtained using various configurations. I am not able to increase the number of points above 64.

    2. Are you using 64 range bins? If so, the I think it's likely you have multiObjBeamForming disabled, when you would like it enabled. See the user's guide for more details.

    Additionally, are you sure the region you're operating in will allow you to do people counting at 77 GHz? I would still highly recommend using an IWR6843 device for this for the additional processing power and because most regulatory bodies are more lenient at 60 GHz.

    Best,

    Nate