08-29-2013 02:59 AM
here is a coffee break solution:
rectify , window peak (crude solution), detect change (and glued some logic to prevent double counts)
won't detect a 'slow' step
08-29-2013 01:46 PM
I found a straghtforward peak detector followed by a transition detector seems to work well with the simulated data.
My initial threshold for the peak detector was zero and that worked but it it seemed to capture one of the phase changes in a way the looked like it might produce errors with noisier data. Similarly the threshold in the transition detector was adjsuted arbitrarily until the results looked good.
Lynn