• State Resolved
  • Locked Locked
  • Replies 5 replies
  • Answers 1 answer
  • Subscribers 30 subscribers
  • Views 508 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

IWR6843ISK-ODS: Architecture of Gesture with Machine Learning demo code

Purvi Mehndiratta
Prodigy 60 points
Part Number: IWR6843ISK-ODS

Hi,

I am running the Gesture with Machine Learning code on IWR6843ISK-ODS. I want to know if the mmwavelib in the sdk is being used in the code (C:\ti\mmwave_sdk_03_06_00_00-LTS\packages\ti\alg\mmwavelib). If so, how is it linked?

What is the architecture of the above mentioned demo code at the highest level of abstraction? 

Regards,

Purvi

  • 0
    TI__Mastermind 21071 points

    Hi Purvi, 

    Looking into this. Please allow 24 hours for an answer here. Thank you for your patience.

    Best Regards,

    Josh

  • 0 in reply to Josh Dye
    TI__Mastermind 21071 points

    Hi Purvi,

    Purvi Mehndiratta said:
    I want to know if the mmwavelib in the sdk is being used in the code (C:\ti\mmwave_sdk_03_06_00_00-LTS\packages\ti\alg\mmwavelib).

    mmwavelib library is not being used in this demo as this library consists of many DSP functions and this demo does not use the DSP. The processing chain used for this gesture demo is a modified version of that which is used in the 6443 Out of Box demo in which only the Cortex-R4f and HWA (Hardware Accelerator) are used.

    Best Regards,

    Josh

  • 0 in reply to Josh Dye
    Prodigy 60 points

    Hi,

    Can you give me an overview of the architectural flow of the code?

    Regards,

    Purvi

  • +1 in reply to Purvi Mehndiratta
    TI__Mastermind 21071 points

    Hi Purvi, 

    Here is a brief overview:

    1. Device boots and initializes drivers, front end, etc... 

    2. Executes hard-coded configuration

    3. Raw ADC data is received from the AFE and processed by the HWA to generate range-doppler heatmap 

    4. Some features are extracted from this data and used to build a feature vector which can be input to the neural network model

    5. Neural network inference is performed based on input values

    6. Feature vectors and results of neural network inference are output via UART

    1 and 2 are only occurring on power up, while 3-6 are repeating every frame. If you have questions about specific parts of the code I would be happy to answer those in more detail.

    Best Regards,

    Josh