Correcting for split-belt treadmill crossover

Hello,

I am using a split-belt treadmill in my lab and am reading z-direction (vertical) forces from it. I use ML to find events such as heel strike and toe off based on where the forces cross a certain threshold.

The problem arises when the subject steps over the midline and on to the other belt, thus not allowing the force on that belt to return to near zero.

I was wondering if anyone had come up with a good solution in their ML code, or perhaps knows enough about this matter to think of one. I have used a couple of different methods, but none have proved to detect every event I am looking for.

Thank you in advance.

Answer 1

[GallVp]

We encountered the same problem in our experiments. We have created an algorithm which we tested in healthy participants. The details were published and a full text is availabe here: Gait event anomaly detection and correction during a split-belt treadmill task

We have also uploaded a MATLAB implementation of the proposed algorithm along with example data. The code and example data are available on Github here: knkTools