Skip to main content
. 2021 Feb 19;12:628728. doi: 10.3389/fpsyg.2021.628728

FIGURE 1.

FIGURE 1

Using body-pose estimation for controlling video stimuli. (A) Representative frames of an example video file showing an actor produce the German Sign Language sign for “psychology” (video courtesy of Henrike Maria Falke, gebaerdenlernen.de; license: CC BY-NC-SA 3.0). (B) Representative frames from the example input video file illustrating the body-pose model which was fit automatically using OpenPose. (C) The information from the fit body-pose model is then used by OpenPoseR to compute the vertical (y-axis) and horizontal (x-axis) velocity of the different points of the model. Based on these calculations, the software then computes the Euclidean norm of the sums of the velocity vectors. (D) The illustrated procedure makes it possible to quantify the total amount of bodily motion in the video using a single measure. Onset and offset the sign are clearly visible as peaks in the plot (shaded areas).