Person sensing / presence detection using CIR

Greetings,
I am working on a full system based on a DWM3000 transceiver which I plan on using for person detection. System is inspired by: Ultra-Wideband (UWB) Radar: Beyond Ranging to Advanced Sensing - Qorvo

The idea is to recognize whether a person is present in the range of this UWB chip (primarily in front, as is illustrated in the webpage above) and I want to do this by analyzing the Channel Impulse Response (CIR). I have 2 DWM3000 chips: one acts as Tx and one Rx. I’ve created a program inspired by SDK example code in which Tx sends a continuous frame and Rx reads the CIR accumulation register which it then sends through serial. By this I got a graph which looks like a correct CIR “waveform”. From my testing, the shape of it seems to change very slightly when I’d put my hand in front, for example. These changes don’t seem to be enough to detect a whole person from a meter away, unless I’m doing this wrong.

My question is, if I could be given some advice on the method to recognize that a person is present in front of the system solely based on analyzing CIR? On top of that I’m wondering what kind of signal I’m supposed to be transmitting during this. Is the continuous frame from the example code enough?

Hello, using 2x DW3000 for radar is a bit tricky.

To get better results, you can for instance increase the preamble length to 1024 symbols, and get the CIR from the ipatov preamble. For a first sample, no real need to use STS.

As a side note, we just released a new kit based on a new chip which provides pretty good radar results out of the box.
The SW release is already out (here: QM35 SDK 3.0 is now available - #4) and the HW kit should be available very soon.

2 Likes

If you only put your hand in front, it is expected the CIR won’t change much. If you stand in front of the system, you should be able to see something.

1 Like