Other Parts Discussed in Thread: IWR6843AOP, IWR6843ISK
Tool/software:
I am considering radar as a means of gathering information for pose estimation in the home.
I am new to the hardware.
What boards and software should I start with?
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.
Tool/software:
I am considering radar as a means of gathering information for pose estimation in the home.
I am new to the hardware.
What boards and software should I start with?
Hello,
IWR Radar by itself outputs processed data such as 1D FFT range profile, point cloud, tracker/cluster algorithms, etc. The DCA1000 is an addon that allows the capture of unprocessed, unfiltered, dumped ADC data directly from the buffer in the form of I/Q real imaginary data. From my experience, we have been able to train models to do basic pose estimation such as standing, lying, and siting using point cloud and tracker alone, which simplifies the process. I would recommend using the IWR6843AOP or IWR6843ISK EVMs for maximum compatibility with our 3D People Tracking example demo which outputs the features needed for pose estimation.
Best Regards,
Pedrhom
Thank you,
Can you point me to the 3D People Tracking example, and can you advise me on a complete list of hardware required? As I say, I am completely new to this.
Regards,
Sean
Hello Sean,
You can find all of our example demos within the Radar Toolbox on the TI Developer Zone. Follow the instructions for installing the radar toolbox via web browser. For the 3D People Tracking demo, here is its user guide. This supplements the code installed with the MMWAVE-SDK-03_06
Best Regards,
Pedrhom
Here is what I will recommend:
MMWAVEICBOOST (optional, this is to use example source code debug tools such as breakpoints and memory browsing)
Radar Toolbox (install via web browser)
3D People Tracking Example Demo
Im guessing you will be using machine learning. For pose, use the tracker's output which includes the track's pos, velocity, and acceleration as features, as well as some of the information about a few points within the track's cluster itself such as position, velocity, SNR.
Best Regards,
Pedrhom
Thank you, that is a huge help.
As you say, I will be using ML, but to do pose estimation, I want to get information about individual body parts and their relationships (angles). Will the hardware pass back a full point cloud? Or just the Doppler information?
Sean
Hello Sean,
The 3D People Tracking Example Demo User Guide covers this. Please check the "UART Output Data Format: section of the guide.
Best Regards,
Pedrhom