I'm trying to figure out if I can use a phone accelerometer (phone is on a belt tightly strapped to low back) to measure sprinting speed. Basically want a cheap option to do feed the cats without having to buy a freelap.
I integrate the z direction acceleration to get velocity (m/s)
And end up with an error in velocity (should be 0 at end). From my research, integrating error in the acceleration signal ends up with a linear drift in velocity. I correct the acceleration by a constant (over duration of sprint) to end up with 0 velocity at the end.
I then do a low pass filter to get rid of some of the noise
When I recorded different gaits (walking through sprinting) I ended up with pretty reasonable speeds that increased as I expected they should
However, I tried recording two different Flys and the first one was much slower than the second. Maybe the first one helped me warm up for the second or is it bad data?
Any accelerometer experts who can tell me if this approach will get me accurate (or at least reliable data)?
Is there a different way I should be correcting the acceleration data besides assuming a linear velocity drift?
Does error increase with acceleration in a consistent manner? For the walking recording I barely had to do any correction