Jason K. Moore

I see. I think we will need both though. Because I hit this "bug" when I was trying to use the mocap column labels to access data in the data...

So this all seems to work. It gives the same results as the Octave file on the test data but when I run it for the plot in our perturbed...

I changed it so I filter the marker positions, compute the derivative, filter the velocities, and compute the derivative instead of filter the marker positions, compute the derivative, and compute...

> I looked in your speedup inverse dynamics branch but could not locate the python code. Pls give a link if possible. inverse dynamics: https://github.com/csu-hmc/GaitAnalysisToolKit/pull/128/files#diff-5e1f4c39e8343cdd4b4dedb9eda0a0efR1084 mimics leg 2d: https://github.com/csu-hmc/GaitAnalysisToolKit/pull/128/files#diff-5e1f4c39e8343cdd4b4dedb9eda0a0efR149 numerical...

> See if my revised Matlab/Octave code helps at all. It eliminates the function call from the for loop. Testing now. > Hope you can track that down. I'll investigate...

> If you run a 2nd order Butterworth filter twice (to eliminate lag), it needs to be designed with a slightly higher cutoff frequency than what you really want. The...

I found some reference here for 2nd order filters: http://nbviewer.ipython.org/github/demotu/BMC/blob/master/notebooks/DataFiltering.ipynb#Butterworth-filter-with-zero-phase-shift Do you know if there is a general formula for any order Butterworth filter?

> That's a very nice notebook. Good to see that he used Pezzack's benchmark data, which was one of the first things I put on the ISB website in 1996....

Ok, I now get virtually identical results with these changes: - add in the filtfilt correction factor - differentiate markers before filtering Left: Python, Right: Octave 

> I assume you're using the conventional digital Butterworth filter Yep, I'm traditional like that :) > which eliminates the need for a C version. I'll implement the C version...