• State
  • Locked Locked
  • Replies 4 replies
  • Subscribers 31 subscribers
  • Views 149 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.

IWRL6432AOP: Evaluation Guide of Presence and Motion Detection with Human vs Non-Human Classification

Jack Cha
Mastermind 7971 points
Part Number: IWRL6432AOP
Other Parts Discussed in Thread: MMWAVE-L-SDK

Tool/software:

Hi 6432AOP Champ

* Missing Document

My customer decided to evaluate L6432 and AOP sensor for People Tracking and Human vs Non-Human Classification feature. 

They tried to download LSDK, then found the following documents. 

They're unable to find Motion and Presence Detection Demo Documentation in the MMWAVE-L-SDK. (i.e. MMWAVE SDK5) at the following location : <MMWAVE_SDK5_INSTALL_DIR>/docs/MotionPresenceDetectionDemo_documentation.pdf

* UART output data format and example

And Using L6432Boost and AOP EVK, They want to test the Human vs Non-Human classification feature using the Presence and Motion Detection Example in Radar Tool code as a baseline. 

Following the classification overview guide, https://dev.ti.com/tirex/explore/node?node=A__ADn5rBEEy5ml6HvCjdb4aA__radar_toolbox__1AslXXD__LATEST

They want to check UART Output Data Format which is introduced in the above document link, 

Can you please let us know how to check the UART output Data ? 

And how can I see the TLV for Human vs Non-Human classification of Detected Object?
Is the below web document snipet related to the following content?
Can you explain with an example of data packet for Detected Object how it is checked in the actual packet?

Customer wants to check the UART Output Data Format for Presence and Motion Detection + Human vs Non-Human Classification Demo.

-------------------------------------------------

Classifier Output TLV

Size: sizeof (tlvHeaderStruct) + numberOfTargets * CLASSIFIER_NUM_CLASSES * sizeof(uint8_t) The Classifier Output TLV consists of an array of classifier outcome in Q7 format, packed as classOutcome[targetIndex][classIndex].

-------------------------------------------------

Thanks and Regards, 

Jack

  • 0
    TI__Prodigy 70 points

    Hi Jack,

    Motion/presence documentation can be found at <MMWAVE_SDK5_INSTALL_DIR>\docs\api_guide_xwrL64xx\MOTION_AND_PRESENCE_DETECTION_DEMO.html. There you can find more detailed info about the demo and output TLVs. Additionally, if you navigate to the above tab labeled 'Algorithms > Classifier > Target Classifier', you can find more info about the classifier itself. The tuning guide (<MMWAVE_SDK5_INSTALL_DIR>/docs/MotionPresenceDetectionDemo_TuningGuide.pdf) is also useful for tuning the demo and better understanding the parameters used.

    In terms of testing Human vs. Non-human classification, running the Industrial Visualizer with the human_non_human_classification config will allow you to see which tracks are classified as human and non-human. A label should appear next to each track with the corresponding classification. If you would like to see the actual UART output from the device, you will most likely have to modify the source code for the visualizer in <radar_toolbox_2_10_00_04_INSTALL_DIR>\tools\visualizers\Applications_Visualizer\common to print the received output.

    I hope that this helps. If not, please clarify how I can better assist.

    Best,

    Yejean Kim

  • 0 in reply to Yejean Kim
    Mastermind 7971 points

    Hi Yejen

    Thanks for your kind reply.

    While I appreciate you pointing me to that documentation, I still don't see any documentation that specifically teaches Human vs Non-Human in TLVs. Neither I nor my clients have been able to find the information we need in those documents, hence this thread.
    I need to know how to tell if a code or TLV value is Human or Non-Human so I can parse it on the host MCU and utilize it in their application.

    As for how to print the UART Data, I will check further and ask again if needed.

    This request is made by the customer to analyze the actual frame structure and design the application to decode the UART data from the host MPU.

    Regards, Jack

  • 0 in reply to Yejean Kim
    Mastermind 7971 points

    As a quick aside, can you tell me which values in the TLVs are used as information to determine Human vs Non-Human? 

    In the image below, are you using the confidenceLevel value from the Target List TLV and using it as the Human vs Non-Human classification value in the Visualizer?

  • 0 in reply to Jack Cha
    TI__Prodigy 70 points

    Hi Jack,

    I think I understand your question now. TLV 317 (MMWDEMO_OUTPUT_EXT_MSG_CLASSIFIER_INFO) outputs a 2D array with two values for each target identified by the radar. One value is the probability of the target being human, and the other value is the probability of the target being non-human. These two probabilities are used to label the track as human vs. non-human. The confidenceLevel value is not used by the classification process. I would recommend looking in the source code for the visualizer to see how it parses the output from this TLV to classify the targets.

    Hope this helps,

    Yejean