I have a UWB radar that Im playing with. The problem is that my algorithms at the moment can clearly, reliably and validly pick up 1-2mm movements of a pendulum i've set up. However, when I use it on my chest to find my heart, things doesnt work. The radar should be set up so that the waves penetrate skin and bone, but I am not entirely sure. The antennae are specced specifically to be able to pick up a heartbeat (or penetrate human tissues).
In terms of math, I first do despiking and a bandpass filter around 0.1 - 5 hz, followed by a principle component analysis to reduce the number of samples/sensors. Then I either do a FFT plus some stuff to find the most pronounced frequencies of the top 5 principle components, or I search for the best fit sinuswave (as the heartbeat can be described as such afaik) to the top 5 principle components.
While these methods give similar results when used on a pendulum, they break down when trying to find the heartbeat. The radar can nicely penetrate wood, and plastic and other stuff. Any ideas?
Bonus question: We had some ultrasound gel at the lab, and i tried to see if removing the "air" component would help, but then all signal dropped. Is ultrasound gel absorbing the radar pulses?