Test that, based on the test file of the dataset https://github.com/moodoki/radical_sdk?tab=readme-ov-file, attempts to detect targets with an indication of their distance and speed.
Download the project to disk:
Download the file with 50 tests from https://fireball.teckyianlim.me/file/flaming-cake/indoor_sample_50.h5
Run the test program:
python Test_Radar_RaDICal_Simplfied.py
As this is a room, to avoid waves reflected by walls and others:
-
Distances are limited by limiting intermediate frecuency
-
Only moving targets are considered. Moving targets are detected by the phase shift between one chirp component and the same component in the previous chirp detected by the same antenna.
Since the formulas to determine the speed that consider the idle_time, adc_sample_rate... etc. have not given me a credible result to determine the speed of the people (which would have to be between 1.11 and 1.66 m/s) and having observed that the phase quotients of two consecutive chirps that have detected an object (a person in this case), are similar to those of another person who had a similar speed (the speed of two people walking in pairs should be similar), I have assumed that the first person detected is going at a speed of 1.2 m/s to determine a factor that is applied to the phase quotient between two consecutive chirps that have detected a person to estimate the speed in subsequent images.
The Texas manual, mmwaveSensing-FMCW-offlineviewing_0 .pdf, is attached. that has been followed in this test, specifically the explanations on page 4 and page 28
References:
https://github.com/itberrios/radical/blob/main/notebooks/range_doppler.ipynb
https://medium.com/@itberrios6/introduction-to-radar-part-2-8a332066917e
Dataset from https://publish.illinois.edu/radicaldata/
https://github.com/moodoki/radical_sdk
https://fireball.teckyianlim.me/file/flaming-cake/indoor_sample_50.h5 Is one small sample (50 frames) to try.
https://www.ti.com/lit/an/swra553a/swra553a.pdf?ts=1714172430750
https://www.plextek.com/articles/a-programmers-introduction-to-processing-imaging-radar-data/ But does not provide dataset for try:
https://www.radartutorial.eu/10.processing/sp04.en.html
https://moodoki.github.io/radical_sdk/ProjectBBoxExample.html It seems to be limited to detecting objects in photos and does not use radar signals. In a similar way to that used in https://github.com/ablanco1950/Radar_Marine-Yolov11 wich uses photos of screen radar. Would be computer vision projects rather than signal processing