• State TI Thinks Resolved
  • Locked Locked
  • Replies 7 replies
  • Answers 1 answer
  • Subscribers 34 subscribers
  • Views 201 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

AWRL6432BOOST: How to increase gesture recognition rate

Batis Chen
Prodigy 90 points
Part Number: AWRL6432BOOST

Hi TI,

I am testing the gesture recognition project with AWRL6432BOOST EVM in toolbox 2.00.00.06.

I found the recognition rate is very low (lower than 10%).

I try to replace the parameter "meanFeat" and "stdFeat" for classifier from 2m to lp in SDK 5.04.00.01 and send the specific chrip ofcurse but the rate still not very good.

Is there any other parameter I can adjust to make the recognition rate higher?

How to generate the values for parapeter "meanFeat" and "stdFeat"?

Is it possible to provide more information for function "classifier_predict"?

  • 0
    TI__Expert 3171 points

    Hello Batis, thanks for your question.  I'll assign this to one of our team members.  In the meantime, could you describe in more detail your setup? for example how close are you to the sensor when you try to do the gesture recognition?  e.g. are you at 2m distance, or less?  Please try to give us a better idea of the actual setup where you get low recognition.  

  • 0 in reply to James Murdock65
    Prodigy 90 points

    Hi James,

    Thanks for reply.

    In the begining, I test it in 2m distance and I got very low recognition rate.

    Then I changed the distaince as 0.5m to 1m in the gesture_recognition_2m_xwrl6432.cfg (sigProcChainCfg2) but no improved.

    Especially "teft to right" is the lowest rate I got.

    Except values in GESTURE_PROBABILITY_THRESHOLDS and GESTURE_COUNT_THRESHOLDS, what else parameter can help to improve the recognition rate?

  • 0 in reply to Batis Chen
    TI__Expert 8782 points

    Hi,

    Which gesture recognition demo are you running? 

    C:\ti\radar_toolbox_2_00_00_06\radar_toolbox_2_00_00_06\source\ti\examples\Gesture_Recognition\xWRLx432_Gesture_Recognition? Are you using the prebuilt binary and default config for that demo?

    Please also confirm the setup that you are 2m from the sensor as shown in the user guide: 

    Regards,

    Tim

  • 0 in reply to Tim Henderson
    Prodigy 90 points

    Hi Tim,

    I am running C:\ti\radar_toolbox_2_00_00_06\radar_toolbox_2_00_00_06\source\ti\examples\Gesture_Recognition\xWRLx432_Gesture_Recognition.

    I am using the prebuilt binary and default config and the distance is 2m.

    Please let me know what parameter I can adjust to improve the recognition rate?

  • 0 in reply to Batis Chen
    TI__Expert 8782 points

    Hi,

    The GESTURE_PROBABILITY_THRESHOLDS and GESTURE_COUNT_THRESHOLDS can be lowered to improve the recognition rate, though I expect this demo to work reasonably well out of the box.

    What is your setting like? Are you in a smaller confined area?

    Regards,

    Tim

  • 0 in reply to Tim Henderson
    Prodigy 90 points

    Hi Tim,

    Is it possible to obtain or compute "distance" information of points in the example code?

  • 0 in reply to Batis Chen
    TI__Expert 8782 points

    Hi,

    In the gesture recognition demo, we use a feature "range average", that takes the average range of the top 25 highest energy points.

    If you want distance from individual points, you should try our one of our point cloud based demos (mmwave_demo, motion and presence demo)

    Regards,

    Tim