Possible collaboration on WiFi sensing

[Zhongqin-Wang]

Hi @ruvnet,

Thanks for your great work on RuView. I have been following the project and found it very interesting.

I am also working on WiFi sensing / ISAC. One issue we have been looking at is how ESP32 AGC and transceiver clock asynchrony affect CSI measurements. From our experiments, these impacts can introduce random phase and amplitude changes of CSI, which may mask the actual human motion we want to sense. This makes precise sensing quite challenging unless they are handled carefully.

Here is one of our related demos: https://www.youtube.com/watch?v=SdnUunW9ueA&t=183s

We have also open-sourced part of the work here: https://github.com/Zhongqin-Wang/Towards-SISO-Bistatic-Sensing-for-ISAC

There are a few other WiFi sensing demos on our channel as well, including single-target tracking, multi-target tracking, breathing / heart-rate estimation, and gesture recognition: https://www.youtube.com/@ZhongqinWang

I would be happy to exchange ideas on ESP32 CSI calibration, AGC and clock asynchrony removal, and robust WiFi sensing if you are interested. Thanks again for the work.

[ruvnet]

I'll take a look.

[ruvnet]

Have you tried this with esp32 csi?

[Zhongqin-Wang]

Have you tried this with esp32 csi?

We have tested our algorithm with ESP32 CSI, as shown here: https://www.youtube.com/watch?v=SdnUunW9ueA&t=183s. It works quite well for precise motion detection, and can also capture respiration and heart-rate related signals from CSI.

The main limitation is the relatively low sampling rate when CSI is streamed out over serial. So we are currently moving most of the signal processing parts onto the ESP32 itself, and only outputting higher-level features such as delay, and micro-Doppler information, or specific delay and Doppler bins' phase. This should also make the data easier to interpret when used with deep learning models.