Because it uses a differences in speed from bin to bin to find a movement onset, it's not agnostic to the bin size. For example, with smaller bins, ds should be smaller because of the smaller timestep, but min_ds doesn't reflect this reliance on timestep.
I think the fix is simply to replace the calls to diff with calls to gradient, but I'm not sure how that would change the rest of the code right now.
Because it uses a differences in speed from bin to bin to find a movement onset, it's not agnostic to the bin size. For example, with smaller bins,
dsshould be smaller because of the smaller timestep, butmin_dsdoesn't reflect this reliance on timestep.I think the fix is simply to replace the calls to
diffwith calls togradient, but I'm not sure how that would change the rest of the code right now.