From 77ea4e84a8203264fa42d4b5a30898ed2489ad9f Mon Sep 17 00:00:00 2001 From: David PAGNON Date: Fri, 2 Sep 2022 11:31:18 +0200 Subject: [PATCH] Shortly elaborate on bone and joint constraints --- Content/paper.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/Content/paper.md b/Content/paper.md index d272a10..cd9bef4 100644 --- a/Content/paper.md +++ b/Content/paper.md @@ -110,7 +110,7 @@ Different filters can be chosen, and their parameters can be adjusted. The user If needed, other standalone tools are provided to further work on the .trc 3D coordinate files (\autoref{fig:utilities}). Among others, it is possible to undersample a file from a higher to a lower framerate, or to convert a file from Z-up to Y-up axis convention. The resulting 3D coordinates can be plotted for verification. Additionally, a tool is provided to detect gait events from point coordinates, according to the equations given by [@Zeni_2008]. ## OpenSim scaling and inverse kinematics -The main contribution of this software is to build a bridge between `OpenPose` and `OpenSim`. The latter allows for much more accurate and robust results [@Pagnon_2022], since it constrains kinematics to an individually scaled and physically accurate skeletal model. +The main contribution of this software is to build a bridge between `OpenPose` and `OpenSim`. The latter allows for much more accurate and robust results [@Pagnon_2022], since it constrains kinematics to an individually scaled and physically accurate skeletal model. Bones are constrained to a constant length, and joints to coherent angle limits. The provided model is adapted from the human gait full-body model [@Rajagopal_2016] and the lifting full-body model [@Beaucage_2019]. The first one has a better definition of the knee joint: abduction/adduction and internal/external rotation angles are constrained to the flexion/extension angle. The latter has a better definition of the spine: each lumbar vertebra is constrained to the next one, which makes it possible for the spine to bend in a coherent way with only a few tracked keypoints, without having to make it a rigid single bone. Combining those two models allows for ours to be as versatile as possible. Hand movements are locked, because the standard `OpenPose` models don't provide any hand detection. This model also takes into account systematic labelling errors in `OpenPose` [@Needham_2021], and offsets model markers as regards true joint centers accordingly. Unlike in marker-based capture, and despite the aforementioned systematic errors, keypoints detection hardly depends on the subject, the operator, nor the context. For this reason, the scaling and the inverse kinematic steps are straightforward, and the provided setup files require little to no adjusting.