This is just an observation, although maybe it has some use. One of my R-Shake units is located near the edge of the road by my house. It detects foot traffic as well as cars. Here is a typical pedestrian walking by, shown both in the time and frequency domain:
This one is a guy walking his dog. His pace is 1.95 steps per second and there is a sharp spike each time the heel of a foot hits the ground. The peak spectral response around 9.8 Hz is the 5th harmonic of the basic stride. The signals from most walkers and joggers look much like this, varying mostly in the pace.
I notice a different signal this morning:
It looks almost upside-down compared with the typical foot traffic. The upper portion of the time-domain signal is a curve rather than a spike, and the spectral peak is at the fundamental. This one was a jogger, with a pace near 3 steps per second but the big difference is that he was running more on the toe and ball of the foot and his heel never quite struck the ground.