It is true that a threshold can be a challenge. Note that I used an adaptive threshold. I took the difference between the maximum and minimum values in the velocity arrays and used 1/3 of that for the threshold. The 1/3 is arbitrary. I tried some other values and that seemed to work well. The velocity is 10 in one data set and 0.3 in the other and this process works for both.
Any time you are trying to automate the analysis of a real (and noisy) signal, you get into situations where you need to look at what you know about the underlying process and what kinds of models and analyses are most likely to work.
For your data some smoothing or elimination of the spikes in the velocity (which almost certainly come from differentiating noise rather than actual velocity changes) might be required.
Lynn
