Ti 6843 mmWave sensor - How can I get micro-Doppler?

Tool: Ti 6843 mmWave sensor

I have created code regarding this radar to collect frames of data. Now what I would like to do is get raw radar data or proccess my current data ( range vs Doppler vs Frames) I really am not sure what to do as my goal is to get a microdoppler plot. Please be as specfic as possible, as I am very new at this.

[William Rose]

@Annie,

What have you tried so far? Please provide the data you want to analyze, and explain its organization, i.e. what the rows and columns are in your data, what the units are, and any other relevant info. Please explain what a microdoppler plot is. In other words, what is on the horizontal and vertical axes of a microdoppler plot? Doing this will help you understand your data, and it will increase your chances of getting useful assistance here.

[Annie]

Micro doppler is -to my knowledge- micro data proccessing on data or raw signals. I'm very much so not an expert. I want to try to adhere to the image below.

This code explains how to collect,store, and present the data.

Here is more in depth explanation of my data -

• RangeDopplerResponse outputs as a 256 by 16 matrix (Power in db)
• RangeGrid outputs as a 1 by 256 matrix (Values of 0 to 12 as a measure of distance)
• DopplerGrid outputs as a 1 by 16 matrix ( Values of -1 to 1 in m/s)
• For a given value from the DopplerGrid and the RangeGrid, a RangeDopplerResponse value is produced.
• I essentially am storing the RangeDopplerResponse over multiple frames. The total data looks like a 3d matrix ( 256*16*Number of frames specified)

At first, I wanted to perform an stft on doppler plot I generated using this code (this would be done on a 2d matrix but there is an issue with putting it in to stft ):

% Define stop time

how_many_frames_to_collect = 200;

% Initialize variables

total_data = [];

frame_times = zeros(1, how_many_frames_to_collect - 1);

frame_collection_times = zeros(1, how_many_frames_to_collect);

for n = 1:how_many_frames_to_collect

[dets, timestamp, meas, overrun] = tiradar();

% meas.RangeDopplerResponse will be empty, if the rangeDopplerHeatMap output is not enabled

% via guimonitor command in config file

if ~isempty(meas.RangeDopplerResponse)

% Store the frame data

total_data = cat(3, total_data, meas.RangeDopplerResponse);

% Measure time difference between frames

if n > 1

frame_times(n - 1) = timestamp - prev_timestamp;

end

prev_timestamp = timestamp;

end

% Calculate collection times

if n > 1

frame_collection_times(n) = frame_collection_times(n - 1) + frame_times(n - 1);

end

% meas.RangeAngleResponse will be empty, if the rangeAzimuthHeatMap output is not enabled

% via guimonitor command in config file

% if ~isempty(meas.RangeAngleResponse)

% rngAngleScope(meas.RangeAngleResponse,meas.RangeGrid',meas.AngleGrid')

% end

% drawnow limitrate;

end

Presenting/grphs of data

total_data = importdata("SIN4.mat")

total_data(total_data > 5) = .8;

% Create coordinate matrices for velocity, range, and frame

[range_indices, doppler_indices, frame_indices] = ndgrid(1:256, 1:16, 1:size(total_data, 3));

range_values = meas.RangeGrid(range_indices);

doppler_values = meas.DopplerGrid(doppler_indices);

%frame_values = frame_collection_times(frame_indices);

%frame_values = frame_collection_times(frame_indices)

% Step 2: Plotting the 3D scatter plot

figure;

h = scatter3(doppler_values(:), range_values(:), frame_indices(:), [], total_data(:), 'filled');

xlabel('Velocity');

ylabel('Range');

zlabel('Frame');

title('Velocity, Range, and Frame');

colorbar;

set(gca,'CLim',[1 2.6]);

average_power = squeeze(mean(total_data, 1));

% Plotting the 2D plot

figure;

imagesc(meas.DopplerGrid, frame_collection_times, average_power');

xlabel('Velocity');

ylabel('Time');

title( 'Time vs Velocity');

colorbar;

set(gca,'CLim',[1 2.5]);

Then I wanted to try something new, performing the proper stft and inverse ffts on the data- method displayed by the image . I'm having issues with performing an stft, as one can't do so on a 3d matrix. Is there a way to do something like this(referring to the image)?

[William Rose]

@Annie, Thank you for the description and code and image. Pardon the length of my comments below - I am writing to learn and remember.

Most of the data is in the structure meas, created by the command above.

Your script collects data with this command, which puts most of the data in structure meas:

[dets, timestamp, meas, overrun] = tiradar();

The command above collects one frame of data, inside a loop that runs as fast as possible. Therefore the interval between frames, when collected as shown in your code, may not be constant.

Your script also reads data from a file:

total_data = importdata("SIN4.mat")

Can you attach SIN4.mat? Does it contain the data returned by tiradar()?

The figure you attached shows doing an FFT along the fast-time dimension, to get the range FFT. The figure then shows doing an FFT along the slow-time dimension to get the Doppler FFT for each frame. It also shows doing a " micro Doppler STFT", apparenty along the frame dimension, to get the STFT. The micro Doppler FFT is shown as an alternative to the Doppler FFT.

Does tiradar() do any of the FFTs illustrated in the figure?

When you provide SIN4.mat, or a file with the data returned by tiradar() (and maybe those are the same thing), we can investigate the possibility of using STFT. If the time between frames is not constant (see above), then we cannot do the STFT on this data. If the time interval from the last data point in frame N to the first data poiont in frame N+1 is greater than the gap between points along slow-time, then we cannot do a normal STFT, but we might be able to do a STFT with careful selection of parameters.

[William Rose]

@Annie,

I looked briefly at this site:

https://www.mathworks.com/help/radar/ug/getting-started-timmwaveradar-example.html

Two examples at that site show tiradar() being called inside a while loop. Your code uses a for loop. The benefit of a while loop is that one can specify a collection duration in seconds, using tic() and toc(). I also see that one specifies the frame rate in a configuration file. I assume you did that. The example recommends a max frame rate of 5 fps.

Structure meas includes meas.RangeDopplerResponse and meas.RangeAngleResponse. I think thse have already had FFTs done to them - but I could be wrong. You have looked at the documentation a lot more than me so you should know. I think the FFTs have been done already because of the coe fragment below:

% Configure and start the sensor by invoking the first step function
tiradar();
% Creates a scope for viewing a range- Angle response map
rngAngleScope = phased.RangeAngleScope;
%  The input data to scope is response data that has already been
%  processed, set IQDataInput as false
rngAngleScope.IQDataInput = false;
% Creates a scope for viewing a range- Doppler response map
rngDopplerScope = phased.RangeDopplerScope;
%  The input data to scope is response data that has already been
%  processed, set IQDataInput as false
rngDopplerScope.IQDataInput = false;
rngDopplerScope.DopplerLabel = 'Doppler velocity (m/s)';

Code above from here. As you no doubt know, raw radar data consists of in-phase and quadrature components ("I-Q"). In fact, if meas.rangeAngleResponse and meas.RangeDopplerResponse are complex, then they have not been FFT's yet; if they are real, they probably have been.

If the data in meas has already been FFT'd, then you cannot do a STFT on it.

If you do

[dets,timestamp,meas,overrun] = tiradar();

from inside a while loop that runs for 40 seconds, with frame rate set to 5 fps, does structure meas contain data for all 200 frames, when the while loop is done? Or is meas getting overwriten on subsequent passes around the while loop? It would help you to know. I recommend doing

whos(meas.RangeAngleResponse)
whos(meas.RangeDopplerResponse)

after the while loop. This will tell you the array sizes, and whether or not the variables are complex.

[Annie]

I am sure the data has already had FFTs done to it. I wonder if I do a inverse fft would that be sufficent and then perform an stft. When I do an ifft I get complex values, but it results in a 3d matrix. I linked an example of my data. Doing what is specified in the document, no matter what stoptime and change of readmode, will always result in a 256 by 16 matrix. This made me wonder if I was only storing one frame or overwriting previous frames, which when graphed and reading the documentation further, supports that assumption. Either way it seems like the previously written code is sufficient for collecting frames of data.

Answer 1

[Honglei Chen]

TI6843 itself does not have capability to output raw data. It only outputs the processed data. Therefore, as is you cannot derive micro-Doppler from the data. I believe the output data is also just the power, thus even IFFT would not help recover the phase (which is needed if we want to recover Doppler or micro-Doppler).

To obtain the raw data, you will need an extra card from TI. See the link below

https://www.ti.com/tool/DCA1000EVM

Hope this helps

[Gayatri Menon]

Hi,

Starting with MATLAB R2024b, the Radar Toolbox Support Package for Texas Instruments® mmWave Radar Sensors allows you to set up and collect raw ADC (IQ) radar data from the Texas Instruments mmWave Radars using the DCA1000EVM capture card.

• Doc link
• Examples
• Supported boards

Hope this helps

Thanks

Gayatri