vvasco
commented
4 years ago This is solved in #280.
To recap how the step length is computed, we filter the distance between the ankles projected on the sagittal plane and extract the maximum values.
The problem was that the distance was not always correct and the filter applied to the ankles signal smoothed too much the signal:
By reducing the depth filtering applied to the skeleton and re-designing the filter applied to the ankles distance, the signal is cleaner:
The videos in #280 show that the step length is higher (~0.5 m) and more reasonable.
In the current simulated TUG, the step length is approximately
20 cm, which seems a bit low. We want to verify this value.