• State
  • Replies 10 replies
  • Subscribers 30 subscribers
  • Views 196 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

IWR6843AOP: The Tracking part of Vital Signs with People Tracking

bl6529
Prodigy 80 points
Part Number: IWR6843AOP

Tool/software:

Hi,

We are working on the Vital Signs with People Tracking demo and want to add a fall detection function on top of it, but there are two problems with the tracking function: (for Wall-mounted radar single-target detection and tracking)
1. Target tracking will always cause the tracking box cube to fly away, resulting in the loss of the target;
2. The tracking result of the radar output target pitch angle and the actual pitch angle are just opposite, when the actual target is below the radar level, the tracking result pitch angle is positive, and the tracking box cube is in a position above the radar level, when the actual target is above the radar level, the pitch angle of the tracking result is negative, and the tracking box cube is in a position that is biased below the radar level.

The above problem caused the fall detection function to not be carried out.

During the experiment, we flashed the "..\mmwave_industrial_toolbox_4_11_0\labs\People_Counting\3D_People_Counting\prebuilt_binaries\3D_people_count_68xx_demo.bin"to test fall detection, the tracking effect of this firmware is relatively stable, may I ask where the tracking problem in vital signs with people tracking is?

My profile for using vital signs with people tracking is given below:

        {"sensorStop \r\n"},
        {"flushCfg \r\n"},
        {"pmicCfg 1 1 \r\n"},
        {"dfeDataOutputMode 1 \r\n"},
        {"channelCfg 15 7 0 \r\n"},
        {"adcCfg 2 1 \r\n"},
        {"adcbufCfg -1 0 1 1 1 \r\n"},
        {"lowPower 0 0 \r\n"},
        {"profileCfg 0 60.75 30.00 1.00 10.00 0 0 200.0 1   96 10785.00 2 1 36  \r\n"},
        {"chirpCfg 0 0 0 0 0 0 0 1 \r\n"},
        {"chirpCfg 1 1 0 0 0 0 0 2 \r\n"},
        {"chirpCfg 2 2 0 0 0 0 0 4 \r\n"},
        {"frameCfg 0 2  48  0 70.00 1 0 \r\n"},
        {"dynamicRACfarCfg -1 4 4 2 2 8 12 4 12 5.00 8.00 0.40 1 1 \r\n"},
        {"staticRACfarCfg -1 6 2 2 2 8 8 6 4 8.00 15.00 0.30 0 0 \r\n"},
        {"dynamicRangeAngleCfg -1 0.75 0.0010 1 0 \r\n"},
        {"dynamic2DAngleCfg -1 1.5 0.0300 1 0 1 0.30 0.85 8.00 \r\n"},
        {"staticRangeAngleCfg -1 0 8 8 \r\n"},
        {"fineMotionCfg -1 1 \r\n"},
        {"bpmCfg -1 0 0 0 \r\n"},
        {"antGeometry0 -1 -1 0 0 -3 -3 -2 -2 -1 -1 0 0 \r\n"},
        {"antGeometry1 -1 0 -1 0 -3 -2 -3 -2 -3 -2 -3 -2 \r\n"},
        {"antPhaseRot 1 -1 1 -1 1 -1 1 -1 1 -1 1 -1 \r\n"},
        {"fovCfg -1 70.0 20.0 \r\n"},
        {"compRangeBiasAndRxChanPhase 0 1 0 -1 0 1 0 -1 0 1 0 -1 0 1 0 -1 0 1 0 -1 0 1 0 -1 0 \r\n"},
        {"staticBoundaryBox -1.5 1.5 0.5 4 0 2.5 \r\n"},
        {"boundaryBox -2 2 0.5 4 0 3 \r\n"},
        {"sensorPosition 1.8 0 15 \r\n"},
        {"gatingParam 3 1 1 2 4 \r\n"},
        {"stateParam 3 3 12 215 50 6000 \r\n"},
        {"allocationParam 20 100 0.1 20 0.5 20 \r\n"},
        {"maxAcceleration 0.1 0.1 0.1 \r\n"},
        {"trackingCfg 1 2 800 1 46 96 70 \r\n"},
        {"presenceBoundaryBox -0.9 0.9 0.5 2 0 3 \r\n"},
        {"vitalsign 15 300 \r\n"},
        {"VSRangeIdxCfg 0 21 \r\n"},
        {"sensorStart \r\n"},
  • 0
    TI__Expert 8600 points

    Hi

    Thanks for your query. Please allow us a couple of days to respond

    Regards

  • 0 in reply to Abhishek Karkisaval
    TI__Expert 8600 points

    Hi

    Vital signs configuration is more tightly coupled to the vitals processing chain and I would not recommend modifying it. Can you add fall detection function with the default Vitals config? In Vitals config, number of ADC samples, frame rate would be such that it is possible to filter out the HR/BR frequency ranges

    Are you able to get consistent tracking when you use the default Vitals config?

    As you have modified the default config from BPM to TDM, you can consider reducing the cfar threshold for better tracking

    Regards

  • 0 in reply to Abhishek Karkisaval
    Prodigy 80 points

    Hi

    Thank you for your reply.

    Abhishek Karkisaval said:
    Can you add fall detection function with the default Vitals config?

    Yes, I'm using the default Vitals config, just modifying the detection range and TDM working mode, which should have no effect on the HR/BR test. By the way, before adding the fall detection algorithm, when I didn't change any configuration parameters, I found that this demo has the problem of unstable tracking, and the target always flies away.

    Is it possible that the "Vital signs with people tracking" demo I got  is not your final version, but an imperfect version in the process, so could you send me a debugged Demo? Thank you for your help.

  • 0 in reply to Abhishek Karkisaval
    Prodigy 80 points

    Hi,

    To add a note, the original demo runs under the default configuration file, and the output tracking situation is unstable, and the pitch angle does not match the actual situation.

  • 0 in reply to bl6529
    TI__Expert 8600 points

    Hi

    I see consistent tracking with the vitals configuration. You can try using the version from radar tool box. You can find the demo at the following location

    radar_toolbox_1_30_01_03\source\ti\examples\Medical\Vital_Signs_With_People_Tracking

    You can use the visualizer from the below location

    radar_toolbox_1_30_01_03\tools\visualizers\Industrial_Visualizer

    Do note that source code is not provided with this demo

    Regards

  • 0 in reply to Abhishek Karkisaval
    Prodigy 80 points

    Hi, thanks for your reply.

    Without the source code, we couldn't do any further development.

    A few months ago we signed the NDA, after that you sent me this version of the source code, I can only further develop on this source code, in order to solve this tracking problem, there are two solutions:

    1).the first is to solve the tracking problem in this source code, I'm not sure if the reason for the unstable tracking is the parameter configuration or the source code logic, so I need your help to see if there is a problem in the source code that you have updated, or how the configuration parameters need to be adjusted;

    2).The second is to send us the source code that you have tweaked.

    If it's just a problem with parameter configuration, it's easier to solve, can you guide me to solve this problem?
    If it is a problem with the logic of the detection or tracking algorithm, it is more complicated and you may need to provide new source code.

    The following file name is the source code package you sent at that time, whitch can help you confirm the version.

  • 0 in reply to bl6529
    TI__Expert 8600 points

    Hi

    The demo in radar_toolbox_1_30_01_03\source\ti\examples\Medical\Vital_Signs_With_People_Tracking corresponds to the source code you already have.

    I did not know the version that was shared with you. There are no more updates on the version you have.

    When you use the visualizer and the config file shared in the location, do you still see that the tracking is inconsistent with the default config?

    Regards

  • 0 in reply to Abhishek Karkisaval
    Prodigy 80 points

    Hi,

    I tried the "radar_toolbox_1_30_01_03" version, the stability of the target tracking will be better, but the pitch angle still seems to be measured inaccurately, and the tracking cube will rise instead when the person is crouching. 

    I compared the configuration file in the "radar_toolbox_1_30_01_03" version and the source version you sent, and there are two differences, one is the frame period, and the other is the fifth parameter "Limit-velocity" in the gating param. As shown in the figure below, the configuration file of the "radar_toolbox_1_30_01_03" version is on the right, and the configuration file in the source code is on the left.

    After I changed the "Limit-velocity" of the configuration file in the source version to 10, the tracking effect is much better, and it looks similar to the radar_toolbox_1_30_01_03 version. Can you explain why? However, the pitch angle measurement is still inaccurate. This has a significant impact on fall detection.Judging from the configuration file "fovCfg -1 70.0 20.0", the pitch angle is only 20 degrees, is it a problem caused by the detection range of the pitch angle is too small? However, judging from the various reference documents, it seems that this angle cannot be adjusted to a large range because of the amount of calculation.

    Hope you can give a solution, thanks!

  • 0 in reply to Abhishek Karkisaval
    Prodigy 80 points

    Hi,

    Additional Notes:

    When I changed the pitch angle in the range parameter "fovCfg -1 70.0 20.0" to 70 and the elevSearchStep in dynamic2DAngleCfg to 3, the radar could not track the target normally, and the point cloud showed a symmetrical state up and down, and a symmetrical point cloud appeared on the roof, Tracking cubes occasionally appear outside of rooftops. which was the same on both versions(radar_toolbox_1_30_01_03 and source code version). As shown in the figure below.

  • 0 in reply to bl6529
    TI__Expert 8600 points

    Hi

    Elevation angle is determined by the antenna pattern. As you cannot change the antenna pattern of the IWR6843AOP, it needs to be set to the default value and any change in the elevation angle would cause tracking failure

    Also, Limit velocity is determined by the frame rate that is used by the configuration. It needs to be set based on the frame rate used by the demo and cannot be arbitrarily changed. For fall detection, you will have to have value for the elevation fft so that you can monitor the variation in height.

    This you can include in the detection chain itself or you can also implement a separate functionality for fall detection where you compute a higher point elevation fft

    Regards