Hello,
First, congratulations on this exciting work!
I had a question about the GRF evaluation described in the paper. The paper states that “Ground reaction force accuracy was quantified as the mean absolute error (MAE) during the stance phase for each directional component, normalized by body weight.” I was wondering how the stance phase was defined/computed for each foot in the LabValidation dataset.
I downloaded the OpenCap LabValidation dataset and tried to identify stance phases using a vertical GRF threshold of 25 N. This works well for the left foot across the walking trials. However, for the right foot in most walking trials, the vertical GRF is already relatively high at (t = 0), which suggests that the right stance phase began before the start of the available time window. In those cases, the right-foot stance phase appears to be truncated.
Could you clarify how these cases were handled when computing stance-phase GRF MAE? For example, were incomplete stance phases excluded, was a different event-detection method used, or were the available windows defined in another way?
I attached a figure showing an example from subject2, walking1, vertical GRF to illustrate what I mean.
Thanks again for the great work!
Hello,
First, congratulations on this exciting work!
I had a question about the GRF evaluation described in the paper. The paper states that “Ground reaction force accuracy was quantified as the mean absolute error (MAE) during the stance phase for each directional component, normalized by body weight.” I was wondering how the stance phase was defined/computed for each foot in the LabValidation dataset.
I downloaded the OpenCap LabValidation dataset and tried to identify stance phases using a vertical GRF threshold of 25 N. This works well for the left foot across the walking trials. However, for the right foot in most walking trials, the vertical GRF is already relatively high at (t = 0), which suggests that the right stance phase began before the start of the available time window. In those cases, the right-foot stance phase appears to be truncated.
Could you clarify how these cases were handled when computing stance-phase GRF MAE? For example, were incomplete stance phases excluded, was a different event-detection method used, or were the available windows defined in another way?
I attached a figure showing an example from subject2, walking1, vertical GRF to illustrate what I mean.
Thanks again for the great work!